jtap-023

download jtap-023

of 53

Transcript of jtap-023

  • 8/8/2019 jtap-023

    1/53

    JTAPJISC Technology Appl ications Programme

    JISC Technology Applications

    A Remote Advisory CaseStudy: The NEAT System

    Mark Ratcliffe

    Tim DaviesUniversity of Wales, Aberystwyth

    Report: 023JISC Technology

    Applications Programme

    Joint I nformation SystemsCommittee

    August 1998

  • 8/8/2019 jtap-023

    2/53

  • 8/8/2019 jtap-023

    3/53

    A Remote Advisory CaseStudy: The NEAT System

    Mark RatcliffeTim Davies

    University of Wales, Aberystwyth

  • 8/8/2019 jtap-023

    4/53

    The JISC Technology Applications Programme is an initiative of the Joint

    Information Systems Committee of the Higher Education Funding Councils.

    For more information contact:

    Tom Franklin

    JTAP Programme Manager

    Computer Building

    University of Manchester

    Manchester

    M13 9PL

    email: [email protected]

    URL: http://www.jtap.ac.uk/

  • 8/8/2019 jtap-023

    5/53

    Abstract

    The Computer Science Department at the University of Wales, Aberystwyth has developed, and is

    operating, an on-line Remote Advisory Service where students are able to obtain personal assistance

    anywhere across the distributed campus without leaving their workstation. This is now being trialed

    by the Open Learning Unit at Aberystwyth prior to its introduction as a service to assist in the

    teaching of their overseas students in locations as remote as Hong Kong.

    The software is part of the NEAT (Networked Expertise, Advice and Tuition) project funded by the

    JISC (Joint Information Services Committee). It has brought together expertise in the field of multi-

    media applications (Department of Psychology, University of Glasgow), distance learning (Open

    Learning Unit at Aberystwyth University) and service providers (Information Services, Aberystwyth

    University).

    The system provides a framework to support new and potentially far-reaching communication methodsfor both one-to-one and group working including mbone video conferencing technology. In addition to

    providing users with a facility to seek assistance from an advisor it also extends the possibilities for

    peer group learning, not only to local students but also those connected over the Internet. Group

    working might be made available to those for whom costs of communication would have previously

    made it prohibitive.

    The two-year project began in 1996 and evaluation has been a central focus throughout. Many

    different types of users in varying situations have used the software and all have provided feedback on

    their experiences.

    This case study describes the findings to date and examines the possibilities that this technology opens

    up.

  • 8/8/2019 jtap-023

    6/53

    2

  • 8/8/2019 jtap-023

    7/53

    Contents:

    1 Acknowledgements.......................................................................................................................... 2

    2 Introduction to the Remote Advisory............................................................................................... 3

    3 Background ...................................................................................................................................... 4

    4 Requirements of the system ............................................................................................................. 55 Stages in the development of the Remote Advisory System............................................................ 6

    5.1 Video and audio transmission................................................................................................... 6

    5.2 Over the shoulder facilities ....................................................................................................... 6

    5.3 Facilitating the connection between User and Advisor ............................................................ 7

    5.4 Developing suitable user interfaces ......................................................................................... 7

    6 NEAT 0: The original Aberystwyth Remote Advisory Service ...................................................... 9

    7 The evaluation process................................................................................................................... 12

    7.1 User interviews ....................................................................................................................... 12

    7.2 The purpose of evaluation....................................................................................................... 13

    7.3 Tour of campus sites ............................................................................................................... 13

    7.4 Open Learning Unit ................................................................................................................ 13

    7.5 Information Services............................................................................................................... 147.6 Language lab facilities ............................................................................................................ 15

    7.7 A demonstration of NEAT 0.................................................................................................. 15

    8 NEAT 1: The current implementation ........................................................................................... 17

    9 Functionality of IP multicast.......................................................................................................... 19

    9.1 Advantages of using IP multicast ........................................................................................... 19

    9.2 Disadvantages of using IP multicast ....................................................................................... 20

    9.3 IP multicast applications of interest to NEAT ........................................................................ 20

    10 How NEAT is using IP multicast ............................................................................................... 23

    10.1 Awareness............................................................................................................................... 23

    10.2 Typical use of NEAT.............................................................................................................. 24

    10.3 Monitoring usage .................................................................................................................... 25

    10.4 Overcoming the limitations of IP multicast ............................................................................ 25

    10.5 Ready for release .................................................................................................................... 27

    11 NEAT 1: The evaluation ............................................................................................................ 28

    11.1 NEAT 1: Its use in Information Services................................................................................ 28

    11.2 NEAT 1: Its use in the Open Learning Unit ........................................................................... 29

    11.3 NEAT 1: The results ............................................................................................................... 30

    12 NEAT 2: The latest implementation........................................................................................... 32

    13 NEAT: Dissemination and publicity .......................................................................................... 33

    14 Conclusion.................................................................................................................................. 35

    15 References .................................................................................................................................. 36

    Appendix 2: Glasgow evaluation team -- interviews with students ...................................................... 37

    1 Computer Science practical session ............................................................................................... 381.1 Students who had used NEAT ................................................................................................ 38

    2 Help Desk Students........................................................................................................................ 42

    3 Open Learning Unit students ......................................................................................................... 45

  • 8/8/2019 jtap-023

    8/53

    1 Acknowledgements

    This report tracks the development of a remote advisory service through various stages of design,

    implementation and evaluation. NEAT (Networked Expertise, Advice and Tuition) is a JTAP funded

    project (JISC Technology Applications Programme) supported by JISC (Joint Information Systems

    Committee)

    1

    .

    The collaborators on the project consist of the Department of Computer Science, Information Services

    and the Open Learning Unit of the University of Wales, Aberystwyth, together with the Multi-Media

    Group of the Department of Psychology from the University of Glasgow.

    Much of the evaluation experience of the collaborators described in this report has been taken from

    internal reports produced by the project.

    The Aberystwyth development team consisted of Mark Ratcliffe (Project Manager), David Price

    (Technical Development), Tim Paul Davies (Implementation Team Leader), Robert Wyn Jones and

    Ivan Jensen.

    The Glasgow Evaluation Team consisted of Jim Mullin, Anne-Marie Fleming, Lucy Smallwood and

    Joyce McLeod.

    The Information Services team consisted of Gill Price and John Chattaway.

    The Open Learning team involved Clare Thomas, John Nelson and Steve Jones.

    The Steering Committee was chaired by Professor Mike Tedd. The committee included Tom Franklin

    (JTAP Project Director), Mike Hopkins (Head of Information Services) and John Martin (University

    of Cardiff).

    1 Funded by the Higher Education Councils of the United Kingdom, JISC is responsible for stimulating and

    enabling the cost-effective exploitation of information systems and provides a high quality national networkinfrastructure for UK higher education and research councils.

    2

  • 8/8/2019 jtap-023

    9/53

    2 Introduction to the Remote Advisory

    In 1995, the Computer Science Department at the University of Wales Aberystwyth began developing

    and operating, an on-line Remote Advisory Service where students were able to obtain assistance

    without leaving their workstation. Students were able to use two workstation rooms on different parts

    of the campus to access an advisor anywhere in the Computer Science Department.

    The Remote Advisory Service not only helped the department provide an efficient service to an ever-

    increasing body of students, but also opened the possibility of providing expertise from outside the

    department, perhaps from other universities or from experts in industry.

    The system provided a framework to support new and potentially far reaching communication

    methods for group working. It extended the possibility for peer group learning, not only to those

    students within computer science, but also to those in other disciplines. Indeed it was thought that

    using this technology, group working could be made available to those for whom costs of

    communication would have previously made it prohibitive.

    The Remote Advisory Service was an adaptation of freely available technology, originally developedfor multicast use (see Section 0), packaged together to provide an integrated facility which provides

    full support of a one:many or many:many advisor to student ratio.

    3

  • 8/8/2019 jtap-023

    10/53

    3 Background

    During the mid 1990s the Computer Science Department at the University of Wales Aberystwyth,

    faced a sizeable increase in student numbers, leading to the opening of a second teaching laboratory

    remote from the main department buildings. The laboratory is equipped with Sun Classic workstations

    running Solaris 2.4, connected to a Sun Sparc 10 as a server. It is fully networked to the rest of thecampus and the JANET network.

    The department emphasises the vocational aspects of computer science and offers a high practical

    content within the syllabus of most modules. Although these practicals are supervised, the department

    has traditionally provided an advisory service, staffed by postgraduate students, to provide assistance

    to students outside the timetabled slots.

    When the new teaching laboratory was opened to support the increasing student numbers, the cost of

    providing a duplicate advisory was prohibitive. It was thus decided to develop a Remote Advisory

    Service which would exploit the multimedia capabilities of the newly installed workstations.

    4

  • 8/8/2019 jtap-023

    11/53

    4 Requirements of the system

    An initial study set out to investigate the requirements of a remote advisory system by first examining

    the existing advisory system.

    Typically, advisors are called in when a student has a problem in compiling a piece of code,difficulties understanding the operating system or has general problems which arise during a session at

    a terminal.

    A problem is usually tackled by the student verbally describing the situation that has arisen. To assist

    in understanding what is required, the advisor usually examines the contents of the relevant window

    (source code, error codes, run time windows, etc.). Once the advisor understands the problem, the

    solution is most often described to the client through reference to the screen; placing the cursor to the

    start of some erroneous code is a typical response.

    It was thought essential that voice communication must be supported and that, to maintain the personal

    nature of the session, a video-link, at least from advisor to student, should be provided. Facilities to

    grab specific windows, look at complete screens and more importantly the ability to share a commonwindow in order to mimic the over the shoulder aspects, were thought to be vital requirements.

    Whiteboard facilities are useful, but are not considered essential.

    In addition to providing the desired facilities for a productive advisory session, it was also deemed

    necessary for the system to have a mechanism which allows students to be queued whilst awaiting

    attention from the advisor. Such an interface would need to give feedback as to the progress of each

    request within the queuing mechanisms and, in the case that there is no advisor logged on at the time

    of the request, a non-interactive support mechanism should be supplied. Such a mechanism would also

    be available to handle less urgent requests for assistance.

    5

  • 8/8/2019 jtap-023

    12/53

    5 Stages in the development of the Remote Advisory System

    Having established the requirements, the following components were identified as being necessary for

    successful development: high quality bi-directional audio communication; uni-directional video

    communication (from advisor to student); a robust and automatic calling and queuing mechanism; and,

    a facility for remote access to the students display. It would also be necessary to design suitable userinterfaces for both students and advisors.

    The communications for a remote advisory service are fundamentally unicast by nature, however,

    given the desire to extend the basic facility to allow support for group working, it was decided to use

    basic multicast tools; multicast can be narrowed to unicast, but not vice versa. The initial system was

    therefore based on: Video Conferencing tool (vic)[1], Video Audio Tool (vat)[2] and WhiteBoard

    (wb)[3]. All of these components are freely available in the public domain.

    After initial considerations, four distinct areas were to be investigated further:

    the effects of differing compression algorithms for video and audio transmission on networktraffic and quality of user service;

    the extension of facilities like wb in order to provide over the shoulder interaction; an implementation of a calling and queuing mechanism;

    the development of a user interface to fit in with other facilities provided by the department(TIPSE[4]).

    5.1 Video and audio transmission

    This study looked at the level of service necessary for both video and audio applications within such a

    system and the resultant influence on bandwidth. Whilst the video might be considered a luxury, it

    was felt to be essential in order to maintain at least some semblance of personal attention on behalf of

    the student. Video traffic can use up large amounts of bandwidth; this work therefore needed to

    investigate ways of providing suitable video quality whilst keeping bandwidth usage to a minimum.

    The implementation of the system would need to consider the fact that it would run not only on thecampus network, but across the Internet in general.

    Audio transmission is far more fundamental to the successful operation of such a system and whilst

    utilising less bandwidth, break-up from high network traffic is much more apparent with audio than

    with video. That is, audio is generally affected by, rather than affects, network loading.

    The video aspect of the study concluded that it was adequate to use only a small frame and that given

    that the transmission would, in normal service, be of the advisor sitting at his or her console, there

    would be comparatively little change between frames. Of the techniques investigated, Cellbs video

    encoding format provided the most suitable service.

    The audio side of the study suggested that besides the actual algorithm used, factors such as type ofmicrophone used and ambient noise levels were significant factors in the quality of audio service

    achieved.

    5.2 Over the shoulder facilities

    This study set out to investigate ways to allow the advisor and student to interact via the screen and

    keyboard as if the advisor was at the students shoulder. Using wb as a starting point, the study

    identified two desirable facilities which were developed for the Remote Advisory Service. These were:

    -

    a screen/window grabber to enable the advisor to "see" what problem the client is discussing;

    a shared interactive workspace, with strict protocol procedures enabling the advisor to take charge

    of whom has control of the workspace.

    6

  • 8/8/2019 jtap-023

    13/53

    The screen/window grabber was introduced as a facility under the advisors control to support screen

    capture and display. It was a tailored implementation ofxwd[5] andxwud. This enabled an advisor to

    see a copy of the students complete display, or just the view of a particular window (assuming that

    the user agreed of course).

    The interactive workspace was adapted fromxkibitz[6]. The advisor who then had control over the

    command line inputs, and was active on the calling students filestore, initiated the call. By using a

    status line in the main window, the student was made aware of who has control of command line

    input.

    By combining these two facilities, the advisor has the ability to see and demonstrate corrections to any

    errors a student may be experiencing.

    5.3 Facilitating the connection between User and Advisor

    All of the above facilities came into operation once the communication had been established between

    the advisor and the student. Prior to this, the main requirement was to provide a queuing mechanism

    for users; this would provide general queuing functions together with facilities to support theadvisor(s) contacting users in the queue. Since the advisors and users are sited in differing locations,

    then this software too would need to consider the remote requirements of the project.

    Initial studies of the requirements of this supporting infrastructure identified the following points:

    as students request assistance, they need to be given details as to how many other people are in thequeue and thus approximately how long they may have to wait before their inquiry is serviced.

    Should there be no advisor on duty, or if the queue is large, then the student should be given the

    option of e-mailing the advisor for future assistance;

    though initially accepting to wait for service, a student may decide to leave the queue. If thisoccurs then the queue must be updated accordingly;

    as the queues are modified, all users and advisors should be given an accurate picture of the stateof the service;

    once an advisor becomes available, the system must inform the student waiting in the queue thatthe advisor is ready to service their request. Facilities must then be provided to make the

    appropriate connection;

    on completion of the advisory session, either side should be capable of terminating the advisorysystem;

    the system must be user friendly; it needs to be implemented in such a manner as to hide systemdetails, for example through the use of default settings for such fields as the name of the machine

    used to host the queue.

    Having established the requirements, the next stage was to decide on a suitable architecture for

    implementation. The need to cater for future enhancements suggested that the best starting point wouldbe to utilise Sun Microsystems Open Network Computing Remote Procedure Calls (RPC) [7].

    5.4 Developing suitable user interfaces

    The development of a user interface for a facility such as the Remote Advisory Service would have

    been a major task in itself. However, in the initial release this task was simplified as the system was

    expected to be an integrated tool within the TIPSE[4], a fully integrated environment providing

    students with many of the software engineering tools used within the department. This project had

    been running for many years and already had a user interface to which new tools were expected to

    conform.

    The TIPSE was written using Tcl/Tk[8], as were the front ends of the basic starting tools ofvic, vat

    and wb, so it followed that the interface for the original remote advisory system would also be written

    in Tcl/Tk.

    7

  • 8/8/2019 jtap-023

    14/53

    8

  • 8/8/2019 jtap-023

    15/53

    6 NEAT 0: The original Aberystwyth Remote Advisory Service

    The original implementation brought together the individual work areas to provide a comprehensive

    system of remote advisory. This represented the starting point of the NEAT project. The software, now

    referred to as NEAT 0, provided two distinct interfaces.

    Figure 1. Students view of the Remote Advisory Service

    The student interface required minimal input to facilitate its use. If there was no advisor in session, the

    student was then prompted to either enter the mail tool or to exit. On entering the mail tool, the student

    only had to type a message or cut/paste output, include a file or a mixture of all three. Addressing the

    advisor was automatic; the student only needed to choose the send button, before exiting the system.

    If the advisor was logged in, the basic Advisory window was displayed on the students screen (Figure

    1). The student then selected the Add to Queue option by clicking the mouse button, and was placed

    on the queue. Details of their position on the queue were displayed to them and updated automatically.

    When the advisor was ready to take their call, the student was notified by a message on the display and

    an unobtrusive audible message via the terminals speaker. Thereafter all audio communication was

    via microphone and headphones, with all volume and transmission levels having been pre-set in a

    start-up file. Termination of an advisory session was instigated by either the student or the advisor.

    The interface to the advisor was rather more detailed (Figure 2) but still retained the same look and

    feel.

    The system was initialised when the first advisor logged in. Any unanswered mail was presented to the

    advisor for response; new mail produced during the session was also directed to this advisor. On start-

    up, the service queue was initialised; as students requested the assistance of an advisor, they were

    added to the queue. In order to minimise the time that a student spent waiting for service, advisors

    were continually informed of the number of students in the queue. Students were also informed of the

    average queuing time before they added themselves to the queue. On accepting a call, the advisor was

    provided with information such as who the student was and which workstation they were using.

    The advisor could invoke any of the window grabbing facilities, shared workspace or whiteboardoptions from their main window. Options for control of input and display variables were made

    available as appropriate within each of the tools.

    9

  • 8/8/2019 jtap-023

    16/53

    Figure 2. Advisors view of the Remote Advisory Service

    The student set (transparently) the authorities necessary for the function of these tools when

    connection was made with an advisor. They were then closed when the session was terminated by

    either party or by failure of the queuing mechanism, for example in the case of the host machine

    failing to respond for some reason.

    Figure 3. A typical Sun screen showing the Remote Advisory Service

    10

  • 8/8/2019 jtap-023

    17/53

    Figure 3 shows a typical layout of a users workspace when working on a Sun workstation. It includes

    a video image of the advisor, shown in the top left corner, the main window (as shown in Figure1), a

    shared workspace and a whiteboard. It is clear that such a layout would be too intrusive on a typical

    PC screen; the average PC does just not have enough real estate to permit such a large area to be

    used by NEAT. Again this was another area to be addressed by NEAT.

    11

  • 8/8/2019 jtap-023

    18/53

    7 The evaluation process

    NEAT has a very distinct difference from that of the original Remote Advisory; it is specifically not

    tailored for Computer Science students. Whilst the original advisory was always intended to be

    generally applicable, it was designed, built and assessed by Computer Science personnel. NEAT on

    the other hand is aimed at a much more general user. We needed to find out these new requirements.

    Evaluation is one of the most important aspects of the NEAT project and is led by the Multi-media

    Communications Group at the Department of Psychology, Glasgow University. The main evaluation

    collaborators are Information Services who provide computing facilities to all Aberystwyth campus

    users and the Open Learning Unit who are the main centre for distance learning courses at

    Aberystwyth.

    One of the major steps in the evaluation of NEAT began when the Glasgow team visited the various

    collaborators in Aberystwyth. This visit took place during the early part of the project and had been

    preceeded by specific videoconferencing interviews to establish the context of use for NEAT in each

    of the evaluation sites.

    The most important aspects of this visit were:

    to meet with users and potential users of the Remote Advisory Service;

    to hold in depth User Interface feedback and discussion meetings face-to-face;

    to promote teamwork between the NEAT project collaborators;

    to look at the physical environment in which the NEAT system was to be used.

    It was interesting to note that although many videoconferences between the collaborators had already

    taken place over the previous few months, the visit seemed to be successful in introducing a more

    cohesive and cooperative relationship between the collaborating sites.

    7.1 User interviews

    During the visit, the Glasgow team interviewed users and potential users from three different groups:

    undergraduate students attending a Computer Science practical lab;

    students who had just used the Help Desk;

    students from the Open Learning Unit who worked at Aberystwyth and therefore could havefrequent access to the Open Learning Unit. They did not consider themselves to be typical of most

    Open Learning Unit students. Nevertheless, the information they gave was of considerable

    importance since their opinions and attitudes did reflect some distance learning related issues and

    could affect the implementation of NEAT within this particular area of education where priorities

    differ from that of full time education.

    Unfortunately the Computer Science users were first years who were potentially less likely to have

    used the remote advisory service (NEAT 0) than other years, and the other two user groups had not yethad a chance to use it. Most of the opinions were therefore from current non-users. The results of these

    interviews are shown in Appendix 1.

    12

  • 8/8/2019 jtap-023

    19/53

    The main negative issues to come out of these interviews were that the remote advisory service was

    not advertised enough, and that some students (in Computer Science) did not know how to open it,

    what they were allowed to use it for, or when the advisors were available. On the positive side,

    practically everyone interviewed was enthusiastic about potential uses of the tool, although some of

    the computer science students did not often feel the need to ask for help, but this is most likely a trait

    of computer science students rather than students in general. Even those who could not see a personal

    use for the help aspects of the tool thought it would be useful for peer communication. Many thought

    that training would be useful, even if it was only an informal introduction to the tool, which was

    especially noted in the Open Learning student interviews. Generally, those who came to the help desk

    were less computer literate than the Computer Science and Open Learning Unit students. The two

    actual users interviewed were very positive about the service they had received so far.

    7.2 The purpose of evaluation

    For evaluation the aim was to bring every stakeholder into the equation and to ensure that, where

    possible, they were considered at all parts of the design and development of the software. Any

    reporting carried out would always look for positive as well as negative aspects, and should any parts

    of the software be critically reviewed, constructive suggestions for solutions to the problem would beprovided.

    During the evaluation, it was noted that the time scale was such that the evaluation of NEAT 1 would

    occur after the final Design Spec for the NEAT 2 version was to be drawn up. This led to appropriate

    changes being applied to the plan!

    The different needs of the Open Learning Unit students compared with full-time on-campus students

    were discussed in terms of type of usage of NEAT; the costs of installing NEAT for students not

    receiving employer sponsorship; and line charges when using a modem. It was clear that the Open

    Learning Unit have a very positive attitude towards the use of computer learning technologies and are

    studying the issues involved in implementing NEAT for their students.

    To keep students informed of NEAT, it was suggested that when students log on, information on

    NEAT should automatically come up. Also, if the NEAT system goes down, users should be told this

    on-line.

    7.3 Tour of campus sites

    The next step in the evaluation process involved visiting the various sites where NEAT was to be

    installed.

    Information Services were planning to install 40 machines across Aberystwyth. That represents 10%

    of the total machines available to students. It was decided that rather than install NEAT on a small

    number of machines in each workstation room, it would only be installed in specialised locations.These included:

    the Thomas Parry Library, situated on the Llanbadarn campus, approximately 25 minutes walkfrom the main campus;

    the Old College, again some 25 minutes walk from the main campus but situated near the centre oftown;

    the School of Art, located between the Old College and the main campus;

    various libraries on campus.

    7.4 Open Learning Unit

    The Open Learning Unit is a department located on the Llanbadarn Campus, which is approximately

    25 minutes walk from the main Aberystwyth campus.It offers courses in: B.Sc. Econ.. Information and Library Studies

    13

  • 8/8/2019 jtap-023

    20/53

    M.Sc. Econ.. Management of Information and Library Services

    M.Sc. Econ.. Schools and Young Peoples Librarianship

    M.Sc. Econ.. Health Information Management

    Undergraduate Certificate in Health Informatics

    Currently, there are over 300 students at the Open Learning Unit, but with new courses, this couldincrease to between 500-600 over the next few years. Currently, Distance Learning students are not

    only spread throughout the UK, but are also located across Europe and in Iceland and Canada. The

    M.Sc. courses are also offered at study centres in Hong Kong.

    The Open Learning Unit currently has 24/25 academic members of staff and 10 support staff who give

    academic advice and pastoral counselling (respectively) to the students. Academic support is given

    through an identifiable academic tutor and through the modular co-ordinator. Students can receive

    pastoral support from a personal tutor.

    As Distance Learning students are mainly in full-time employment, the course modules are written on

    the basis of 15 hours of study per week. The three year course is made up of 23/24 modules (equal to

    240 credits) and 4 examination components. Courses can last from 3-5 years.

    Ideally, students require co-operation from their employers, however support varies as about a third of

    the course are fully sponsored by their employers; a third are partially sponsored; and the remaining

    third receive no support from their employers at all. Some students in the latter category do not inform

    their employers that they are studying on the course because of the negative attitudes that employers

    sometimes have towards employees who are endeavouring to improve their career position through

    study. This means that not only are some students disadvantaged financially, but they may also be

    disadvantaged in terms of the non-availability of study time during working hours or by difficulties in

    study motivation in terms of career development.

    All students on the B.Sc. Econ.. Information and Library Studies must have access to PCs and use

    Microsoft Works. Since January 97, the latest B.Sc. Econ.. Information and Library Studies course

    uses First Class computer conferencing software for extra support. Students are mainly female and

    mature students; they would need some technical support.

    Distance Learning students visit the Open Learning Unit three times in the course of their study, to

    attend a Study School held at the beginning of each new term year. They may also visit the Open

    Learning Unit at the beginning of their 3rd

    Year for dissertation information. Study School also gives

    them the opportunity of meeting with other students on the course and perhaps forming peer support

    networks with telephone or email support. Distance Learning students can also set up regional

    meetings - currently there are three groups in Scotland.

    The Open Learning Unit can obviously make good use of NEAT. Exactly how the evaluation is to becarried out is described insection 0.

    7.5 Information Services

    Evaluation of NEAT by Information Services was to concentrate mainly on the help desk that serves

    all staff and students at Aberystwyth. In time, NEAT might also be used within the library section of

    Information Services. The help desk was located within the computing science building at the start of

    the project but moved to the Hugh Owen building before the trials began.

    The Information Services help desk provides help to anyone who has a problem, either by email or in

    person. It covers a variety of different problems, the most common of which are:

    loss of password;

    HTML queries and finding certain web resources;

    UNIX queries - can be setting privileges for web pages;

    14

  • 8/8/2019 jtap-023

    21/53

    unusual statistics queries - can be UNIX;

    closing user accounts for administration staff (e.g. if a user has not paid fees).

    After examining a number of sessions, the evaluators identified the following points:

    a lot of the solutions were written down on postit notes and given to the customer to take away

    with them; the checking account status is done on a character based terminal;

    the problematic package is always opened to recreate the problem and demonstrate how to dealwith it;

    sometimes the customer is asked to demonstrate how they would go about the problem task;

    the advisor often points to the screen when discussing the subject (it would be useful in the remotesystem for the advisor to be able to point with the mouse and have a pointer show up on the users

    machine.);

    information is often sent to the customer by email.

    All of these points fed into the requirements phases of NEAT 1.

    7.6 Language lab facilities

    In addition to the main collaborators, there is one further area that had potential attraction to the NEAT

    project. The European Language Department has two language labs in operation, each with about

    twenty machines. One is slightly older and has 386s, but is open 24 hours. The other room contains

    very new PCs and runs Windows NT. This room is only open certain hours. At beginner level,

    students would be expected to spend 1 hr per week in the labs, and later about 2 to 6 hours in the lab.

    There are taught classes for most of the working day held in the labs. Students can access foreign

    programs over satellite or cable. They currently have 100 Mbit connections, which allows them to see

    films from the server. They will soon have a Video-on-Demand system so that students can watch

    foreign films.

    Potential uses of NEAT for the European Languages department were identified as: distance lecturing;

    one:one communication;

    link with penpals in foreign country;

    broadcast locally between the machines in the lab;

    group translation of passages using different colours to mark text.

    For NEAT to be usable by the language students, sound quality must be good and preferably provide

    lip synchronising of the video and audio.

    Multimedia does not support all the functions of the language lab and teaching, but it can be useful for

    people who are unwilling to talk during a one hour conversation class (the British can be quite shy,

    and dont like to expose themselves in front of their peers). It will also be useful for people to log in

    from home and get extra support and facilities. Students should come out of the course with the skills

    of working together using the equipment, and the skills of using video technology.

    7.7 A demonstration of NEAT 0

    In order for the evaluation team to fully appreciate NEAT, a demonstration session was set up in the

    Computer Science Advisory (advisor) and in the Solarium workstation room (users). Richard Huss (a

    post-graduate student working on the advisory desk) gave advice to users that allowed the evaluators

    to observe the Advisory system in situ. This was followed by a visit to the Solarium where 2 users of

    NEAT were observed.

    The following observations were made at the advisor station: the advisor logs on to the system and any email messages appear;

    15

  • 8/8/2019 jtap-023

    22/53

    clicked on Connect Next and as it was starting said This takes a long time;

    whiteboard - fairly slow to come up and took a while to send data; when it did - no feedback;

    Xterm - fairly slow to come up (Xkibitz comes up first); screening down - window goes fromblack to white to black - can be distracting;

    status bar did not show any messages at first;

    the video took a long time to appear after the status bar said Established audio and video. Whileit may be true that the connection was set up, this message is incorrect from the users perspective

    until they can see the evidence of this;

    the advisor moved the whiteboard window, and covered up the status bar;

    imported text into Xterm OK;

    screen dump - status bar message - OK, but took a long time before there was any feedback thatthe screen dump was happening;

    clicking on the Xterm button to close it again looked unintuitive (similar to the way that SDRlooks at sessions);

    advice was given using both audio and text;

    noted that background conversations in the room which is shared with others, seemed to disturbthe advisor even though he was using headphones i.e. sometimes had difficulty in hearing user;

    audio level higher so that audience could hear user through headphones;

    while the advisor was serving the first user he did not notice that there was another user in thequeue. The information telling him that someone had joined the queue was not clear enough or

    obvious enough. Because the advisor didnt disconnect the user immediately after finishing, the

    other user was kept waiting longer than necessary, although this may have been due to the

    presence of observers;

    windows should come in the correct place so that people not confident with managing windowsare not forced into moving things (Experience shows that some people will leave a window hidden

    and not use it, even if it is needed, rather than ask what they should do);

    disconnecting the users takes away the tools so that the user does not have time to write down orsave the information the advisor has given you (this was commented on by the user). This equates

    to someone on the help desk writing some information down on a postit note for you, and thenputting it in the bin and not letting you take it away. It also comes with no warning, as the user has

    no control over the disconnection if the advisor decides to send him away. The user would not

    know beforehand that the tools would all disappear on disconnection.

    The following observations were made at the user station:

    a first year student had logged on to the service, however there was a loss of audio at the user end,though not at the Advisors end. This could have been due to the need to adjust the sound level

    through the Audio Manager. Possibly a more experienced user could have resolved this

    problem. Troubleshooting this sort of problem where the audio would have to be tested would be

    very difficult unless a user was very confident. As a result of the one-way loss of audio, both user

    and Advisor used the whiteboard. The user did not know how to enter text on the whiteboard;

    one of the NEAT team demonstrated the user end on another machine. The audio managersound level was adjusted. Initially, audio came through the speakers rather than the headphones

    which seemed to disturb students at the nearby workstations who stopped working to look at the

    user.

    16

  • 8/8/2019 jtap-023

    23/53

    8 NEAT 1: The current implementation

    Having spent two years working on the original version of the Remote Advisory Service, we knew

    that there were many areas for improvement, indeed some of these points have already been addressed

    in this report. One of the changes thought most important by certain user groups was to maintain

    attributes on the abilities/ experiences of advisors; in such cases it was felt that a student should beprovided with enough information about an advisor in order to select the one most appropriate. This

    could have been achieved in the prototype by extending the queuing mechanism to support multiple

    advisors (many queues each attributed with the advisors and their skills). However, such a system

    hosted on a single machine would neither be robust enough nor extensible enough to be of general use.

    Not only would the advisory fail if the host machine went down, but extending the service to a number

    of different departments would lead to an unmanageable number of queues.

    The infrastructure of the original Remote Advisory Service was just not extensible enough to be

    practical for wider scale application. It also became clear that the collaborators working on NEAT had

    requirements, which although not addressed by NEAT itself, were of such a similar nature that it

    would have been foolhardy to have designed a system restricted to the original infrastructure. At the

    very least, it was thought that the system should consider the following possibilities: Passive Advisor: a potential advisor might monitor the service and only become active if loads

    on the existing active advisors reach an unacceptable level. This does not have to be restricted

    solely to advisors on call; it might also, for example, be of particular use to managers of an

    advisory service for monitoring usage.

    Remote Tutoring: there is no reason why a session should be restricted to one:one. Indeed,students might have been discussing a problem with each other before one of them calls up the

    Remote Advisor for assistance; all of these students might be interested in hearing what the

    advisor has to say.

    Group working: students might like to use the service for contacting members of their group inorder to hold a group meeting. The system might provide support by broadcasting information

    about the group meeting. Other members might then elect to join in the remote session.

    Two design requirements thus became apparent. Firstly, a single advisory session may contain more

    than two individuals. This confirmed our original ideas that it would be best if the media used to

    support the session (e.g. video, audio, whiteboard, etc.) were transmitted in a multicast manner.

    Secondly, the individuals who might make use of NEAT need a mechanism to gain awareness of

    which experts and students are around and whether or not they are free. Figure 4 depicts how such a

    network might appear. It shows the various categories of user that might be available and the flow of

    communications that might be set up to support them. The diagram shows how Active Advisor 1 has

    been connected to Student 2. Group 1 has two students communicating, as does Group 2, and there is a

    Passive Advisor monitoring the state of communications over the network. As yet, Active Advisor 2

    and Student 1 remain unconnected.

    17

  • 8/8/2019 jtap-023

    24/53

    Group 1

    Student 1

    Group 2

    Student 2

    Student

    1

    Passive

    Advisor

    Group 2

    Student 1

    Student

    2

    Group 1Student 2

    Active

    Advisor

    1

    LAN/WAN

    Flow of

    Communications

    Key:

    Active

    Advisor2

    Figure 4. Users communicating over the network

    In order to address these issues, a new approach was designed based on the group awareness work

    undertaken by Ben Anderson [9] now at British Telecom.

    The main idea behind this new approach is that packets of information are multicast around the

    network using a Group Awareness Protocol (GAP). Packets transmitted from an advisor include

    information on their skills, together with attributes such as the length of their queue (which may be

    empty). An awareness tool operated by a student wishing to contact an advisor, listens out for advisor

    packets in order to build up information on the skills that are currently available. Having chosen the

    most promising of the available advisors (both in terms of skills and queue size), the students

    awareness tool uses other information broadcast in the advisor packets, to set up a session with the

    chosen advisor. This information includes details on the advisors hardware/software configuration as

    well as information on the advisors location.

    Exactly the same protocols can be used to support group working; in this case a student broadcasts

    information equivalent to My name is Fred; I would like a group meeting. Other group members

    might be listening in to the network and when the appropriate packets are received, they can set up

    their conference. A similar scenario can be used to set up multi-way conversations between groups of

    advisors and students. Of course such working arrangements might have a severe impact on network

    traffic; this is an area that will be examined carefully during the evaluation of NEAT.

    18

  • 8/8/2019 jtap-023

    25/53

    9 Functionality of IP multicast

    Having established the requirements of the NEAT system, and produced a design which would

    support the necessary infrastructure, the next step was to examine how these ideas might be

    implemented to provide the necessary communications. This led to an investigation into various data

    communications protocols.

    In designing a network-based application we had to decide which communications protocols to use.

    TCP/IP (Transmission Control Protocol/ Internet Protocol), which we will go on to discuss, is the most

    widely used set of protocols for computer communication. Hence, we decided to use this as it would

    open NEAT up to as many users as possible. TCP/IP is a suite of protocols that interconnects the

    various different systems and network types on the Internet and allows for communications across

    these heterogeneous networks. Three protocols are most commonly used within the TCP/IP scheme:

    IP, TCP and UDP (User Datagram Protocol). IP lies at the centre of the TCP/IP protocol suite and all

    data on the Internet flows through IP packets. Higher level protocols, such as UDP and TCP are built

    upon this. We became interested in IP multicast, an extension of UDP, because it easily allowed for

    one-to-one, one-to-many and many-to-many group communications. While TCP is more widely used

    in network applications, UDP, and hence IP multicast, is a lower-overhead alternative. UDP is oftencompared to a mail service where individual messages are sent (like letters or postcards), while TCP is

    compared to the telephone service, where an actual connection is established between the two

    communicating parties.

    An investigation into UDP revealed that:

    it is unreliable. UDP has no mechanisms for detecting errors or re-transmitting corrupted or lostdata. Moreover, packets may arrive out of sequence. TCP, on the other hand, detects and

    retransmits corrupted or lost segments (the unit of TCP transmissions). Although TCP is more

    reliable, there is an overhead associated with ensuring its reliability;

    it is connection-less. With UDP there is no negotiation of a connection before transmission. TCPsets up a connection before transmission. This means that UDP is often faster;

    it is message-orientated. Messages are packaged into individual datagrams (packets) while TCPallows for a seemingly continuous stream of data. UDP is therefore suitable if the information

    sent can be seen logically as discrete messages.

    While unicast involves sending packets from one host to another and broadcast entails sending a

    packet to all hosts on a given subnet, multicast involves the transmission of IP datagram packets to a

    host group. The fact that UDP is connection-less and message-orientated makes multicast possible. It

    would be impossible with TCP because a connection would have to be established with each group

    member. IP multicast involves:

    sending a single copy of a message to the group as a whole, not one copy per receiver. Multiplecopies of the message are only created when the paths to group members diverge;

    the group is defined as one or more hosts identified by a single IP address (known as IP class Daddresses in the range 224.0.0.0 to 239.255.255.255);

    a multicast groups membership is dynamic. Hosts can join and leave at any time. Moreover, thereare no restrictions on the number or location of host group members and a host may be a member

    of more than one group.

    9.1 Advantages of using IP multicast

    TCP is much more widely used than UDP by network applications, so what would be the advantages

    of using multicast communications in an application like NEAT? We have identified the following as

    important:

    it easily allows for one-to-one, one-to-many and many-to-many communications which means that

    we can support not only one-to-one advisory but also group working and remote lecturing. NEATis written in Java which allows for easy-to-use platform-independent and cross-platform multicast

    communications;

    19

  • 8/8/2019 jtap-023

    26/53

    senders do not have to know who the receivers are. This means, for NEAT, that users of the system(both advisors and students) simply multicast their presence. Exactly who receives the information

    is unimportant to the sender. Also, at the application level, receivers only have to join the multicast

    group to receive and are not required to register or connect with some central server. This allows

    for the decentralisation of networked-based applications. Our first prototype Remote Advisory

    Service had one central queuing/ administration server through which all communications weremade [10]. The decentralised design allows for a more robust (as there is no reliance on any one

    host) and dynamic system;

    using multicast leads to reduced bandwidth usage when compared to TCP connection-orientated orbroadcasting modes of communication. As for TCP, a connection is not required with each

    receiver, and for broadcasting, uninterested hosts are not sent messages. Moreover, an increase in

    the number of receivers is not met by a commensurate increase in bandwidth use. Senders dont

    have to send more packets when the number of receivers increases. This means that many more

    users of a multicast-based network application could be supported at any one time for the same use

    of bandwidth. An implication of this is that multicast is more suitable for bandwidth-intensive

    applications, such as real-time video and audio, across the internet. These are the very capabilities

    that NEAT requires;

    communications can be geographically scoped. The Time To Live (TTL) on a multicast packetrefers to how many mrouter hops the packet can make. Mrouters decrement the TTL value and then

    forward the packet until its TTL has reached 0 when the packet is discarded. This geographic

    scoping could allow advisory session communications to be constrained to one site (a university

    campus LAN for example).

    9.2 Disadvantages of using IP multicast

    Using IP multicast has its limitations and disadvantages. We have identified the following as

    problems:

    UDP is, by its very nature, unreliable; a sender cannot be sure whether a packet that was sent wasreceived at all, whether its data was corrupted or, whether packets were received out of order. UDP

    places error detection and correction at the application level; address - port space. A multicast session is defined by a multicast address and port number

    combination. Clashes of addresses and/ or ports do sometimes cause disruption to multicast

    applications;

    although most Ethernet LANs support multicast over the same physical network, many systems arenot configured to send and receive multicast traffic. To send multicast over WANs there must be a

    system of multicast-enabled routers (mrouters) between remote sender(s) and receiver(s). This

    system of mrouters is known as the Multicast BackbONE (MBone).

    9.3 IP multicast applications of interest to NEAT

    On balance, we felt the advantages of multicast outweighed its disadvantages, particularly as many of

    its limitations presented solutions (discussed later). Moreover, there are also a number of public

    domain multicast tools for collaborative working that we could make use of. Having established thatmulticast would be the best route to support NEAT, it soon became clear that the number of multicast-

    enabled applications that are available is growing rapidly. A decision was made very early on in

    NEAT that the project should not attempt to develop individual tools for collaborative work - it would

    instead provide the infrastructure necessary to support the communication between users. NEAT has

    been designed such that new tools can be plugged into this framework. This section looks at a number

    of multicast tools and protocols that are available in the public domain and that have been considered

    useful for NEAT, either as plug-ins or in the design of the infrastructure.

    9.3.1 Applications

    We investigated a number of IP multicast applications and high level protocols for their suitability insupporting the requirements of remote advisory.

    20

  • 8/8/2019 jtap-023

    27/53

    The following applications were considered important:

    Video: The Video Conference tool (VIC) provides effective real-time video transmission andreception [1]. Advantages of using VIC were that its settings could be configured and its GUI could

    be customised as its interface is fairly technical (see Figure 5). Versions of VIC run on UNIX,

    Windows 95 and NT4 and are cross-platform compatible.

    Figure 5. VIC 2.8 before (left) and after (right) customisation.

    Audio: The Video Audio Tool (VAT), is a multicast audio transmitter/ receiver. VAT could also beconfigured and customised (see Figure 6). RAT, Robust Audio Tool, is another audio tool that

    allows multiple users to talk over the MBone but is also more adaptive to network and host

    conditions [11,13]. However, RAT cannot be easily customised and thus is of less use to NEAT

    with its requirement for a simple user interface. Versions of VAT are also multi- and cross-platformcapable.

    Figure 6. VAT 4.0 before (left) and after (right) customisation.

    21

  • 8/8/2019 jtap-023

    28/53

    Whiteboarding: Whiteboards are shared drawing spaces also used to export, view and annotatepostscript files. A number of whiteboards were investigated, namely: wb, mDesk and TeleDraw

    [12,13 and 14].

    Shared text editing: NTE is a shared text editor that allows multiple users to edit a single textdocument [15]. Each user may then save copies of the document. Most importantly, versions of

    NTE run on UNIX, Windows 95 and Windows NT4 and between these platforms. Shared Applications: Many of the NEAT collaborators felt that some form of generic shared

    application tool that allowed multiple users to view and participate in a particular application (word

    processing, spreadsheet, database etc.) would provide important functionality to a remote advisory

    service. No such tools currently exist that use multicast. For one-to-one advisory sessions, we have

    investigated commercial software, in particular Symantecs pcANYWHERE which allows

    advisors to view another users screen and to take control of their PC [16]. This is very useful for

    the one-to-one advisory scenario. We have successfully plugged this tool into the PC version of

    NEAT 1. Another area of our research is concerned with cross-platform application sharing.

    9.3.2 Protocols

    The following protocols were considered important in building the communications infrastructure to

    support remote advisory: Conference creation, scheduling and announcement protocols such as the Session Description

    protocol (SDP), Session Announcement Protocol (SAP) and Session Invitation Protocol (SIP)

    [17,18 and 19]. The session directory tool (SDR) [20] was investigated as it provided much of the

    video conference session description generation, announcement and invitation facilities that NEAT

    could use. SDR advertises multicast conferences, and provides a listing of conferences being

    advertised. SDR uses SDP to provide sufficient details for other users to elect whether to join a

    given advertised session and to facilitate actually joining the session (i.e. to start the right tools

    with the correct configurations). SAP is used to publicise session descriptions, usually by

    multicasting SAP datagram packets on a well-known address and port. SIP is used to invite

    individual hosts to a session on a one-to-one basis. It was considered important that NEAT should

    use standard protocols for compatibility reasons as well as to avoid re-inventing the wheel if theycould help meet our requirements.

    Group Awareness Protocol (GAP) [9]. Teleport, an awareness tool, allows other users of the tool tosee whos about and what they are doing [21]. GAP is used to multicast information about each

    Teleport user to everyone else. From this, Teleport users can initiate conversations with others

    using video and audio tools.

    22

  • 8/8/2019 jtap-023

    29/53

    10 How NEAT is using IP multicast

    NEAT itself provides the infrastructure for remote advisory: the sending and receiving of awareness

    information, the capability to call an advisor, a queuing mechanism associated with each advisor, the

    creation and joining of sessions and the hiding of the technical details from the user. The first release

    of NEAT 1 has the multicast tools VIC, VAT and NT plugged in but also uses pcANYWHERE toenable the advisor to view the students screen and to gain control over that students PC.

    NEAT 1 has two classes of user: advisors and students. NEAT users all multicast awareness

    information on a configurable address-port combination using the Group Awareness Protocol (GAP).

    Advisory services offered by different departments may be provided on other address-port

    combinations. For example, Computer Science advisory could be run on address/port

    225.2.13.127/20000 while European Languages advisory could be run on 225.7.13.100/22002. This

    has the advantage that users of each advisory service (both advisors and students) are unaware of the

    existence of the other service and its users.

    10.1 Awareness

    The periodically multicasted awareness packets provide information on which users are available andwhat they are currently doing. To achieve this, each user sends out information on some, if not all, of

    the following areas; some of this information is gained from the NEAT system itself, while the rest is

    obtained from the user, either the first time they use the system or while the system is running:

    personal details: this includes the users name, position (e.g. lecturer, student, manager etc.),department (e.g. Physics, Personnel etc.), institution (university or business name), subject (course

    student is taking) and a students year;

    contact details: e-mail address, phone number(s) and URL contact details;

    attributes: areas of expertise/ skills for advisors; other attributes could include languages spoken,level of computer literacy (for students) etc.

    state information: is the user/ advisor busy, how many people are waiting in an advisors queue,

    which advisor has a student that has requested help (if any), what conferencing tools are currentlybeing run on which addresses/ ports and with what other parameters; and what session(s) are they

    presently involved in;

    host/ site details: this includes the users id, the operating system they are running (e.g. Windows95), the machines domain name and IP address, and the local time and other locale-specific

    details;

    capabilities: what bandwidth limitations the user faces (e.g. it could be 33.6 kbps for a modemuser), hardware available (i.e. do they have microphone, speakers and/ or a video camera), what

    software plug-ins they have for NEAT (e.g. do they have NTE a shared text editor);

    and, messages to the whole group: an advisor might want to tell everyone Im off for a coffee,back by 11.20.

    By listening in to the awareness packets, every user of NEAT is able to build up a table of all otherNEAT users based on their latest awareness information. This technique provides up-to-date

    information to each user on all of the above areas about every other user running NEAT. This table is

    used for a number of purposes:

    providing each user with awareness about what else is happening reduces the chances of confusionand more easily facilitates the development of new working practices when using the tool. This sort

    of information is often lacking in other tools;

    it gives students all the information they need to select an advisor most suited to discussing theirproblem;

    it provides advisors with detailed information on students seeking their advice. This enablesadvisors to be better prepared to solve users problems;

    it is used by the NEAT system itself for error detection, correction and recovery based on the statesof other users;

    23

  • 8/8/2019 jtap-023

    30/53

    it enables the NEAT system to create session descriptions based not only on the requirements of theproblem but also on the specifics of each users hardware, software and bandwidth capabilities and

    each users personal requirements, preferences or needs;

    it contains information on all sessions that are currently running and can therefore be used to enableother students or advisors to request to join a session or be invited into a session already running.

    This allows for the dynamic membership of sessions that are more than simply one-to-one.E.g. ifan advisor is in session with a student and they cant solve their problem or answer their request,

    they could invite another advisor in to the conversation to help;

    this information is used by the Reflector/ Gateway, when non-multicast-enabled hosts are involvedin the advisory service. In this way it is known which media streams need to be turned to and from

    multicast and to and from which users IP addresses.

    10.2 Typical use of NEAT

    Typically, the system will be used as follows: a student experiences a problem or wishes to make a

    query while seated at his/ her PC. The student launches the NEAT remote advisory system which then

    starts periodically sending awareness information out about the student as well as receiving

    information about everyone else (using GAP). In this way it is able to build up a table. Each entry in

    the table contains the latest awareness information from each user running NEAT on the samemulticast address and port combination. The student can then select an advisor and, if required, obtain

    more information on him/ her.

    Figure 7. Students screen while connected showing NEAT (top right), the video image of the advisor

    (VIC) and the audio controls (VAT). In this example the student is also working on a document using

    Word.

    24

  • 8/8/2019 jtap-023

    31/53

    Once a suitable advisor is selected, a message is sent to the advisor and the student is then placed on

    their queue. When the advisor is ready to communicate, the advisors NEAT software generates a

    session description (using SDP) containing information on which tools to start and with what

    parameters. This is sent to the student using the Session Invitation Protocol (SIP) and is then used by

    the students NEAT system to join the session and launch the tools specified by the advisor. The

    student and advisor are now in communication (see Figs. 7,8). The advisor may invoke the most

    appropriate tools based on the nature of the students request. When the advisor has resolved the

    students query, the connection is closed and the advisor can go on to deal with the next student in

    their queue.

    Figure 8. Advisors screen while connected showing NEAT (top right), video controls (VIC), audio

    controls (VAT) and pcANYWHERE (left) now controlling the students screen.

    10.3 Monitoring usage

    Our use of multicast communications and GAP makes it fairly straightforward to develop a monitorfor the system. Such an application simply has to listen on the same multicast group address and port

    combination that GAP packets are being sent out on and then build up a detailed log of the systems

    usage.

    10.4 Overcoming the limitations of IP multicast

    There are three main limitations in using IP multicast: communicating with non-multicast-enabled

    hosts; unreliability; and group address/ port generation. We have considered the first to be the most

    serious concern for NEAT as solving it opens up the system to many more people who for various

    reasons (including recent trends toward switched ethernets) are not using multicast. However, we have

    also considered the other two, less critical, concerns.

    25

  • 8/8/2019 jtap-023

    32/53

    10.4.1 Communicating with non-multicast-enabled hosts

    Because not all potential users of NEAT can send and receive multicast, we have built a Gateway/

    Reflector (see Figure 9) to convert multicast traffic into unicast and vice-versa (both awareness traffic

    and the media traffic such as video and audio). Using the reflector means that non-multicast-enabled

    hosts can effectively join multicast groups. Unicast traffic from non-multicast-enabled hosts is

    forwarded to other non-multicast-enabled hosts and multicasted. Multicast traffic is turned into unicastand forwarded to each non-multicast-enabled host.

    To the users of NEAT these implementation details are hidden. Instead of multicasting their awareness

    packets, non-enabled hosts send them using unicast UDP directly to the Reflector/ Gateway. Whether

    the user is able to use multicast or not does not affect the functionality of the system itself, these

    details are hidden from the user. The Gateway/ Reflector knows whats going on by looking at

    awareness packets and builds up a table similar to that of the users of NEAT. From this it can tell what

    unicast traffic to multicast and forward and what multicast traffic to forward. This means that non-

    multicast-enabled hosts can dynamically join and leave multicast groups as if they were on the

    MBone. Obviously this doesnt come with the other advantages of being on the MBone, most

    importantly it is not bandwidth efficient. In summary, the Gateway/ Reflector allows for group

    communications among any combination of enabled and non-multicast-enabled hosts (includingentirely amongst non-enabled hosts) and gives access to NEAT to sites that would be otherwise off-

    limits.

    Figure 9. The Reflector/ Gateway converts multicast traffic to unicast and visa versa

    10.4.2 Unreliability

    While multicast is unreliable in the sense that packets can get lost, can arrive out of sequence or

    become corrupted, these factors are not critical to the NEAT infrastructure itself, which is able to

    recover from these sorts of errors. It does become more important, however, for the tools that NEAT

    uses. We are currently investigating the Reliable Multicast Protocol (RMP) that provides mechanisms

    for acknowledgement, resending of lost packets, ordering and repair/ request policies. RMP involves

    building reliable mechanisms on top of IP multicast. A RMP could well open the door to a wider

    variety of multicast applications that require a high level of reliability; this could include shared

    application or remote control-like applications (similar to pcANYWHERE) that currently use TCP.

    26

  • 8/8/2019 jtap-023

    33/53

    10.4.3 Multicast address-port conflicts

    Finally, multicast address/ port conflicts are another area of unreliability with regards to multicast

    applications. It is virtually impossible to guarantee that a tool can generate a unique address/ port for a

    session. This is because there are no ways of reserving addresses, and address generation is

    decentralised (i.e. addresses are generated by individual applications such as SDR, or NEAT and are

    not requested from some form of central server). When tools are launched on the same address, errorscan occur or confusion can arise. We have investigated a number of ways of reducing the chance of

    address/ port clashes. These could be used in isolation or combination:

    before generating a session description, the NEAT system could examine the table of latestawareness information from each user and see what addresses/ ports are in use. It can then know

    not to use any of these addresses. This virtually eliminates chances of a clash with anyone else on

    the same NEAT advisory service;

    NEAT could listen for a session announcement from tools like SDR that generates multicastannouncements using SAP on a well-known port and address. It could then avoid using any

    addresses in each announcements session description. However, there is no guarantee that people

    will announce sessions before using multicast tools;

    NEAT could listen on an address for a while and after a predefined period clear of any traffic, deem

    the address to be unused and make use of it. However, how long should one wait before thinking itunused? There is still no guarantee that someone else will start using it later or that the address was

    in use but at the time of listening there was no traffic.

    Although the chances of a clash are minimal, the results could prove disastrous to an advisory session,

    so while attempts at minimising the risk are costly in terms of effort, they are useful.

    10.5 Ready for release

    NEAT 1 produced an efficient information-rich system that provided an open and dynamic interaction

    framework using MBone technology. Having implemented and fully tested the system, the next stage

    was to determine user reaction.

    27

  • 8/8/2019 jtap-023

    34/53

    11 NEAT 1: The evaluation

    NEAT 1 was released to the project on schedule in mid-July 1997. As each installation was slightly

    different due to collaborator requirements and site specifics, the first installation carried out was for

    Information Services at Aberystwyth.

    11.1 NEAT 1: Its use in Information Services

    Information Services began installation of NEAT during the summer of 1997 on to approximately 40

    different machines distributed across campus. Prior to the installation it was announced that the whole

    campus was being upgraded from Windows 95 to Windows NT. This was a fortunate development

    for NEAT since testing of the system had shown that NT was a much more reliable platform than

    Windows 95. Testing had consistently shown that the NEAT software, though written for 95, was

    running much better on NT.

    A few months earlier, Information Services had requested that the NEAT development team add

    functionality to the toolset to better meet the needs of their users. Unfortunately although the decision

    to concentrate on the infrastructure still looked as though it had been the correct thing to do, by the

    summer of 1997 the tools had not progressed as much as expected (though this was to change radicallyby late 1997). A decision was therefore made to incorporate pcANYWHERE into the toolset.

    pcANYWHERE is a commercial product that enables an advisor to fully interact with a users PC and

    this includes any applications that might be running on the users machine. It costs 17 per user

    machine and was purchased by Information Services for all of their trial machines. Minor

    modifications were made to NEAT to support this software.

    To install NEAT, each machine was upgraded with a sound card (not installed as standard at

    Aberystwyth) and a pcANYWHERE licence. The sound cards were to create all sorts of problems for

    the project. Early requirement analysis, undertaken by the project as a whole, had shown a need for

    full duplex communications. Most sound cards available provided the necessary hardware and many

    drivers were now supporting full duplex. Although tests undertaken by the development team had

    found a number of satisfactory cards, there were serious problems with common cards such as the

    Soundblaster (though later releases of the drivers were to address this problem). As far as Information

    Services were concerned, as they were installing new sound cards, the need to install specific sound

    cards presented no difficulty.

    The development team produced a full installation tool (for NT and 95) based on the Freeman

    Installer. It was intended that this tool would provide an automated installer for use by a nave user.

    The installation within Information Services proved very successful though they were hampered by

    technical problems caused by the cross-campus upgrade to Windows NT.

    A workshop on how to use NEAT (from an advisors perspective) was held in early September. It was

    directed specifically at the needs of Information Services. Approximately 10 people attended all ofwho were potential advisors within Information Services.

    Information Services have provided full NEAT service in the form of Help Desk Live 9 to 5 since

    October 1997. Numerous advertising campaigns to promote the service have been undertaken and

    although successful, usage has not been as high as expected.

    Further discussion on this issue follows in section 0.

    28

  • 8/8/2019 jtap-023

    35/53

    Evaluation team at Glasgow

    Once the installation had been completed by Information Services at Aberystwyth, work began in

    installing NEAT at Glasgow. This work was undertaken by Lucy Smallwood and Joyce McLeod of

    the Multimedia Communications Group. The intention was to run through the whole installation

    without over-the-shoulder assistance from Aberystwyth. This would enable the evaluation team to

    ascertain how easy other users would find the installation process.

    Unfortunately the installation did not go well. The biggest problem was that it was to be installed on

    Windows95, which the development team knew was more troublesome. Reports from Glasgow

    interpreted by the development team, suggested that the sound card drivers were causing the problem.

    This turned out to be the case, though there were further problems caused by incompatibility of the

    camera drivers.

    This delay was disappointing to the project, not least because of the wasted effort by Glasgow, but an

    idea of using the Faculty of Art at Glasgow as a trial site also fell through. This was an NT site and

    likely to have generated many useful users. The problems that Glasgow had with installing NEAT on

    95 was probably detrimental but the requirement for full duplex sound cards provided a major

    stumbling block.

    Further discussion on this issue follows in section 0.

    11.2 NEAT 1: Its use in the Open Learning Unit

    Even before the release of NEAT 1, the Open Learning Unit had undertaken an investigation to

    establish how the concept of a remote advisory service could be used to support distance learning.

    This involved two different threads. On the one hand, experiments were undertaken with modems to

    evaluate the performance of NEAT to establish the working limits of the tools. These experiments

    involved detailed analysis of bandwidth required by NEAT and its toolset. The results of these

    experiments are discussed in section 0. While these experiments were underway, a study by the

    Glasgow evaluation team was carried out to interview potential students.

    The results of the studies were used to formulate the way in which the Open Learning Unit might use

    NEAT. Although the exact format has yet to be finalised, discussions, both internally and with the

    Glasgow evaluation team, have led to the following scenarios being considered:

    Pastoral tutorials;

    Pre-assignment advice;

    Essay and assignment feedback;

    Dissertation discussions;

    Group support sessions.

    The experimental trials of NEAT are scheduled to take place between February and the end of the

    academic year. It is likely that each student will be offered up to four NEAT sessions during thisperiod. Each session will have to be pre-arranged for a suitably convenient time, and will last for a

    probable maximum of 30 minutes. After each session, both the staff and students will fill out a

    questionnaire.

    The Open Learning Unit is currently contacting their students inviting them to participate in the trials.

    They have been given a full specification of the hardware that NEAT requires and are offered all of

    the software required for the trials. The Open Learning Unit have also offered to reimburse the cost of

    the phone calls made during the trials.

    The results of these trials will be published during the summer of 1998.

    29

  • 8/8/2019 jtap-023

    36/53

    11.3 NEAT 1: The results

    NEAT 1 was a great success; it provided all the functionality that was expected of it and was in use in

    a variety of different locations. A large number of presentations were given both on a national and

    international forum; the responses were very positive. Unfortunately our experiences with its

    application were not so encouraging.

    By early December 1997, there had been a number of setbacks, in particular:

    the progress report from Glasgow was very negative; they had still not installed NEAT 1 correctlydespite positive experience elsewhere in the project;

    the tests with modems using the traditional multicast tools had been discouraging;

    Information Services were receiving only a couple of enquiries per day on the NEAT 1 system;

    Within Computer Science, the Remote Advisory System (NEAT 0) was hardly being used.

    11.3.1 Technical issues

    Glasgow is a Windows95 site. This in itself should not have been a problem; a number of trials at

    sites remote from Aberystwyth were also on 95 machines and all these tests had been successful. The

    sound cards were a definite problem area; we were convinced that the drivers for the sound cards were

    still causing problems, particularly when used by vat. It soon became apparent that in all cases, it was

    the tools, not the NEAT infrastructure, which was providing the difficulties.

    What we needed was a different approach to the tools; vic and vat were causing problems with the

    sound cards, pcANYWHERE was causing problems with licensing costs. Fortunately a solution was at

    hand; recent investigations into Microsoft NetMeeting (available free to anyone with a Microsoft O/S),

    looked very promising. By late December 1997, as a result of work undertaken as a final year project

    in Aberystwyth, the development team had ascertained that NetMeeting:

    * could be configured so that it would work with NEAT; it could be customised to be initiatedfrom the command line, hiding all details from the user. The user interface could be configured to

    be very similar to NEAT 1;

    * could replace vic/ vat and pcANYWHERE; tests indicated that vic and vat were not as efficientas commercially developed software. NetMeeting provides all the required functionality of

    pcANYWHERE without the additional cost;

    * provides better 1:1 audio and video quality;* reduces the sound card problem; the tests carried out with NetMeeting were much more

    successful than with other software with which we had experience.

    Initial tests demonstrated that NetMeeting would reduce restrictions and increase functionality. After

    further experimenting, it was decided that NetMeeting would be customised and plugged into NEAT

    as a replacement for vic, vat and pcANYWHERE. The UNIX version would retain vic and vat until a

    port of NetMeeting was available for UNIX platforms.

    When the move to NetMeeting was first considered, the Glasgow evaluation team expressed concernthat such fundamental changes to the software would affect their evaluation process. They thought that

    this would involve lots more work, and the lessons from the previous version could not be carried

    forward. In adopting NetMeeting it was decided that although the tools might change, the functionality

    must remain the same. There would be no great difference from the users point of view. The NEAT

    interface itself would not be significantly changed.

    Most important to the project was the fact that NetMeeting had proved successful over modem

    connections, with 2-way video and full-duplex audio being achieved over a 33.6kbps modem. This

    change meant that full trials would at last be possible for the Open Learning Unit, which until now had

    been dogged by bandwidth problems.

    30

  • 8/8/2019 jtap-023

    37/53

    11.3.2 Usage

    Investigations by the collaborators indicated that the poor usage of NEAT1 was due to the set up

    within Information Services and the Department of Computer Science. The Glasgow Evaluation team

    had already identified such problems when evaluating NEAT1.

    Information Services have NEAT installed on some 40 machines, however this is less than 10% of thetotal. One has to bear in mind that each installation required a sound card (18) and a pcANYWHERE

    license (17). In this respect, achieving a total of 40 installations was very encouraging. However, as

    there is such a high demand for all machines within Information Services it was not practical to

    reserve machines for NEAT usage. In any case, NEAT was not intended to be used in this way. The

    student should be able to use NEAT to gain access to an advisor at any point during a session. Without

    the student even having to leave the workstation, an advisor should be able to capture the screen or

    interact with the application to solve the problem. NEAT was not intended for use as a dedicated

    machine to be used only when a problem arises.

    This problem cannot be solved overnight but by replacing pcANYWHERE with NetMeeting, the

    installation costs of NEAT are reduced quite considerably. Even more promising is that the

    management of Information Services has recognised the real potential that NEAT has to offer. It istherefore very likely that Information Services will increase the availability of NEAT quite

    dramatically.

    Within Computer Science the usage of NEAT 0 by undergraduate students was thought to be low

    mainly because of the close proximity of the workstations to the advisors. To visit the advisors in

    person involves a walk of less than 50m which has meant that students can just as easily seek the

    advisor in person as use NEAT.

    31

  • 8/8/2019 jtap-023

    3