DL:Lesson 5 Classification Schemas Luca Dini [email protected]
Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA...
Transcript of Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA...
![Page 1: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/1.jpg)
PANEL SOFTENG
Software Challenges for New
WWW.IARIA.ORG
1
Petre DINI2015BARCELONA
Software Challenges for NewServices/Devices
Petre DINIConcordia University, Canada
China Space Agency Center, China
IARIA Organization
![Page 2: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/2.jpg)
New Paradigms
• Smart cities
• Smart devices
• Smart technologies
• eGovernment online services
Urban computing, apps, embedded software
222Petre DINI
2015BARCELONA
• Approximate computing [Big Data]
• Eventual consistency [Replicated Data]
Storage/computation, data centers management
• Programming paradigms [agent-, context-, network-coding]
• Designing/development paradigms [agile, crowdsourcing]
• Machine learning [knowledge, semantic, taxonomies,]
• etc.
![Page 3: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/3.jpg)
Changes in Technology/Service Lifecycle
Agile methods
Open source
Crowdsourcing
Design security/privacy-driven
333Petre DINI
2015BARCELONA
Design security/privacy-driven
Design for accessibility
Adaptive interfaces
….
![Page 4: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/4.jpg)
Technology/Maturity Lifecycle
NoSQL
maturity
smart phonesapps
cloud ~10 years
cloud: data-as-a-servicebandwidth-on-demandutility-on-demand……
fog
cyberspacesoftware
444Petre DINI
2015BARCELONA
idea convincing marketenthusiasm
realityvalley
getting commercial
waterfall
crowdsourcing
fogcomputing
time
![Page 5: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/5.jpg)
Today’s Panelists
• Moderator:Petre Dini, Concordia University, Canada | China Space AgencyCenter, China
• PanelistsLaura Semini, Università di Pisa, Italy
> software product lines
555Petre DINI
2015BARCELONA
> software product lines
• Michael Jäger, Technische Hochschule Mittelhessen, Germany
> security- and privacy-aware systems …software manipulations
• Stepan Orlov, St. Petersburg State Polytechnic University, Russian
Federation
> scaling the server, using clouds, and client-side computing
![Page 6: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/6.jpg)
Q & A
Qs & As
666Petre DINI
2015BARCELONA
WWW.IARIA.ORG
Qs & As
![Page 7: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/7.jpg)
Opportunities for non-commercial web applications
to use computational
S. Orlov, N. Shabrov
SOFTENG 2015, Barcelona, Spain
to use computationalresources
St. Petersburg Polytecnical University
Computer Technology in Engineering Dept.
http://equares.ctmech.ru
![Page 8: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/8.jpg)
Background information
• More and more software is designedin the form of Web applicationsaccessed through a regular Webbrowser
• Simulation software is not an
2
• Simulation software is not anexception
• https://simscale.com• http://www.wolfram.com/mathematica• http://www.wolframalpha.com/• https://insightmaker.com/
• We've just added our own• http://equares.ctmech.ru/
commercial, runssimulations on server
non-commercial,simulations on client
non-commercial,simulations on server
![Page 9: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/9.jpg)
Background information
• Why yet another simulation software?• Something important is missing in other tools
• What we need• Ability to formulate algorithm for numerical experiment in a flexible
way• Efficiency
3
• Efficiency• Considerably faster, than MATLAB/SciLab• Ability to avoid extra memory usage in algorithms
• Accessibility• Web-application running in browser
• Ease of use• No algorithm programming. Instead, draw a scheme of data flows between
data processing units• Ability to use an existing scheme and slightly modify it if necessary
• Extensibility• Open source code• Everyone can add missing features that he/she needs
![Page 10: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/10.jpg)
Background information
• Example 1 – double pendulum• There are quasi-periodic and
chaotic motions• To see that, one should build the
Poincare map• But it’s too difficult in
• MATLAB, SciLab
4
• MATLAB, SciLab• Mathematica (before version 9)• Maple
• In any system (except XPP)programming is necessary, andsometimes the solver has to bere-implemented
• And it will work too slow
![Page 11: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/11.jpg)
Background information
• Example 2 – Mathieu equation
• Ince – Strutt diagram showsparametric resonance instabilityareas on plane of parameters
5
areas on plane of parameters• How student can obtain it?
• Analytically – too difficult• Numerically – well, the fundamental
matrix has to be computed at eachpixel, and its eigenvalues – a bit toodifficult
• And it will work too slow if we useMATLAB/SciLab
![Page 12: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/12.jpg)
Choice of technologies
• Server• Node.js (web-server)
• С++, Qt (computational core running on server)
• MongoDB (server database)
6
MongoDB (server database)
• Jade (HTML generation engine)
• Client• HTML 5 (using SSE)
• D3 (interactive visualization of schemes)
• MathJax (formulas in documentation)
• CSS, JavaScript, jQuery, jQueryUI (common things)
![Page 13: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/13.jpg)
Simulation specification
• There is a scheme consisting of boxes and links (pipes-and-filtersdesign pattern)
• Links determine data flows• Boxes are data processors• A box has ports; there are input and output ports• A link connects an output port of a box and an input port of some
other box
7
other box• Boxes without input ports are data sources• Boxes without output ports are data storage• A box can have parameters
![Page 14: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/14.jpg)
Simulation specification
• Simulation scheme defines adata processing systemimplementing numerical experiment
• User draws simulation scheme in browser and
8
• User draws simulation scheme in browser andsends it to server as JSON
![Page 15: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/15.jpg)
• Simple pendulum, one phase curve
Examples
9
![Page 16: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/16.jpg)
Examples
• Simple pendulum, phase volume evolution
10
![Page 17: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/17.jpg)
Examples
• Simple pendulum, skeletal curve
11
![Page 18: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/18.jpg)
Examples
• Interactive phase portrait
12
![Page 19: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/19.jpg)
Examples
• Ince – Strutt diagram (500 x 500, 6.3 s)
13
![Page 20: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/20.jpg)
• Double pendulum, Poincare map• 50000 points, 28.5 s
Examples
14
![Page 21: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/21.jpg)
Examples
• Strange attractor• Forced Duffing oscillator
• Interactive parameter change
15
![Page 22: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/22.jpg)
Examples
• Colored Mandelbrot set• Interactive pan/zoom
16
![Page 23: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/23.jpg)
Examples
• Chaos in logistic mapping
17
![Page 24: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/24.jpg)
• Set of points where a function has a positive value• http://habrahabr.ru/post/135344/
Examples
18
![Page 25: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/25.jpg)
Scalability problem
• Simulations run on server
• Each client can consume a thread for as long asuser wants
• Some simulations finish fast
19
• Some simulations finish fast
• Some run for a long time
• Currently web server is one 16-cpu node
• What if our it will become popular?
![Page 26: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/26.jpg)
Scalability problem
• Possible solutions• Add more servers
• Who will buy them?
• Limit user resources on server• Limit memory – ok
20
• Limit memory – ok
• Limit CPU time – won't help much(user may really need ~1 minute, or maybe more)
• Move simulations to client• Back to desktop applications?
• Use JavaScript or asm.js? But we're using LAPACK and will addother 3rd party libs
• Cloud computing• Who will pay? Or can it be available for free?
![Page 27: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/27.jpg)
Thank you!
21
Questions?
http://equares.ctmech.ru
![Page 28: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/28.jpg)
Software Product Lines
&
Feature Oriented Systems
Behaviour analysisLaura Semini
Dipartimento di Informatica
Università di Pisa, Italia
![Page 29: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/29.jpg)
• A Software Product Line is a family of software products that:• share a common architecture • varies in the functionalities offered
• Product line engineering accelerates product development by
• Exploiting the commonalities among the product line members while• Managing the differences (called variabilities)
• The definition of models to specify and analyse Software Product Lines is a current challenge
• potentially exponential number of different products that can be obtained
![Page 30: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/30.jpg)
-oriented development
• Much of today software is developed using scenarios or use cases
• Essentially, a use case is the specification of a feature
• Feature = Incremental unit of functionality [def by Pamela Zave]
• Moreover, features are a convenient way to model
• commonalities and variabilities of Software Product Lines
![Page 31: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/31.jpg)
-oriented systems: interactions
• Features may interact in unexpected ways• Initially addressed in telecommunication systems
• Features1. are often engineered by people who are not aware of the feature-interaction
problem, and 2. are used in safety-critical domains.
• Therefore:• (intended) feature interactions must be very well explained, and • we should not strive for solutions (for unintended interactions) that work only half
the time.
![Page 32: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/32.jpg)
Feature diagrams
![Page 33: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/33.jpg)
:
design vs development perspectives
• Two essentially different perspectives on feature interaction:
1. Design:
Here the focus is on understanding requirements with a view towards building a system in which features either do not interact, or do so in a positive way.
2. Development:
The focus here is on code, with a view towards analysing and understanding an existing system.
![Page 34: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/34.jpg)
Features:
annotation vs composition- based
• Composition-based
• Composition of behaviours
• Annotation-based
• Implemented with directives to thepreprocessor
e.g. M. Jackson and P. Zave.
Distributed feature composition:
a virtual architecture for telecommunications services. IEEE TSE 1998
![Page 35: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/35.jpg)
challenges
• Model behaviour in feature-based systems.
• Look for patterns of feature interaction.
• Designing a “feature-aware computational model”
![Page 36: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/36.jpg)
Trusted Ubiquitious Computing
Michael JägerTechnische Hochschule Mittelhessen
SOFTENG, April 2015
Jäger, THM Trusted Ubiquitious Computing SOFTENG’15
![Page 37: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/37.jpg)
A User-centric View on Ubiquitious Computing
Smartphone
Smart Car Smart TV
Smart WatchSmart Home
User
Control
Information
machine-to-machine communication
Third Parties
Information
Control
Smart devices- have cameras- have microphones- are location-aware- provide private information- are remotely controllable- are exposed to - misuse - internet attacks
Behind the scenes:
Critical Infrastructures- provide - communication, - positioning, - traffic control, - etc.- are exposed to internet attacks.
Can I trust my personal systems?
Can I trust the infrastructurebehind my systems?
- Who is observing me?- Who is controlling my devices?- Is it safe to use my systems? ...
Systems Engineering Challenge:How can we build trustworthy systems?
Jäger, THM Trusted Ubiquitious Computing SOFTENG’15
![Page 38: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/38.jpg)
What is Trust?
An entity can be trusted if it always behaves in the expected manner for the intended purpose(Trusted Computing Group).
Claims
Trust is a key factor for acceptance.
Misuse and manipulation will happen increasingly.
How to build trusted systems?
Security by Design
I Take cyber-attack and information misuse attempts for granted.I Model misuse cases, derive security requirements.I Be concerned about privacy.
Use trusted Components
I Secure network protocolsI Open source softwareI Manipulation-protected key components: "Trusted Device"
Jäger, THM Trusted Ubiquitious Computing SOFTENG’15
![Page 39: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/39.jpg)
Trusted Information Agent (TIA): Technical building blocksEnhanced Android Platform
Extended Dalvik VM
Java TPM Access Library
(JTSS)
TPM Access Library
IF-MAP
Client Library
for Android
Linux Kernel
IF-MAP Server
Trusted Network
Connect
/ IF-MAP
Remote
Attestation (TPM)
Chain of
Trust (IMA)
Trusted Computing
Group Standards
Kernel Module
uses
API forMetadataExchange
OpenSourceImpl.
conforms to
Early TCG Standard
Part of TNC Standard
(Coppolino, Jäger, Kuntze, Rieke: A Trusted Information Agent for Security Information and Event Management, ICONS 2012)
Jäger, THM Trusted Ubiquitious Computing SOFTENG’15
![Page 40: Software Challenges for New Services/DevicesSoftware Challenges for New 1 Petre DINI 2015 BARCELONA Services/Devices Petre DINI ... • Machine learning ... Agile methods Open source](https://reader033.fdocuments.in/reader033/viewer/2022042022/5e79b86b6e2de66ae930d122/html5/thumbnails/40.jpg)
TIA architecture
GPSTime
AuthorityPrivacy CA
Evidence
Generator
IF–MAP
Server
State
Visualiser
Location Trusted Time
Network State
AIKTrusted Time
Signed Evidence
AIK
(Coppolino, Jäger, Kuntze, Rieke: A Trusted Information Agent for Security Information and Event Management, ICONS 2012)
Jäger, THM Trusted Ubiquitious Computing SOFTENG’15