An Experimental Testbed for Building Service-Based Media Applications

(Outline of paper draft to be submitted in TridentCom 2010 as testbed practices.Authors: Sang Woo Han, Namgon Kim, Bumhyun Baek, and JongWon Kim)

Sang Woo Han

swhan@gist.ac.kr

Networked Media Lab.,

Dept. of Information and Communications

Gwangju Institute of Science and Technology

Lab seminar (2009-10-31)

Outline

• Discussion points in this talk– What are practical examples of media-oriented service composition

experiment (towards immersive future media services)?• Multimedia Service Composition, Visual Sharing, IMS, RealCast for realistic

broadcasting, ...?– What is the proper term that expresses immersive future media services

based on composable services? (In this talk, the term “Service-Based Media Applications” is used temporally.)

• Problem statement– Designing testbed platform, mainly focusing on experiment control

• Problem definition– OMX Toolkit development, integrating Java-based software agents,

TB management tools, virtualization tools, and measurement tools.• Solution

– TBD

Towards Immersive Future Media Services

• Personal needs about immersive media services are growing.

• Using widely dispersed media contents and tools, flexible media systems are necessary.

• Service orientation is a driving force to instantaneously and flexibly build value-added multimedia applications on top of composable services in the Internet.

Referring to “Why do we need a content-centric future internet?” created by the Future Content Networks Group, May 2009.

Service-Based Media Applications by Service Composition Paradigm

• A media application reuses individual media services to form new and complex process in a time-/cost-aware manner.

• A media service is a functional entity that assists processing and communication of media streams.

Towards Service-Oriented Testbed for Media Experiment Support

• Dynamic service composition by applications– Dynamically compose services depending on the functional

relationship among services and handle events of testbed

• Service and resource adaptation combined with monitoring– Monitor resource utilization and adapt services accordingly

– Dynamically adapt resources according to the requests of a service

Construct an extended testbed supporting complicated service composition and stable operation through extended tools for service control and management

Background• Designing service-oriented testbeds needs to consider several aspects:

– On the testbed designed for supporting media processing and delivery and testbed control framework to operate the testbed,

– We need service control framework to assist experimenters in development and testing of service composition methodologies by using testbed capabilities

• Composing individual services according to dynamic context changing

• Guiding the design, development, deployment, operation and retirement of services delivered by a service provider

• Integrating heterogeneous resources and providing them

TM Forum SDFs (Service Delivery Frameworks) supports the efficient design, creation, deployment, provisioning and management of seamless services across different access networks

TEMPO: NICTA Service Delivery and Testbed Framework

• Service delivery framework– Support for multiple type of services (e.g., VoD, multiplayer online gaming)

– Secure and managed P2P approach

– Optimize distributed service delivery in terms of energy consumption

• Unified testbed framework (OMF)– Control, measure, and manage federations of networking testbeds

Fraunhofer FOKUS Open SOA Telco Playground to enable dynamic service composition for multiple network technologies.

Service Tools for Composition/Provision/Monitoring

• Gush– Application development environment– Installation on remote hosts– Software installation – Execution and monitoring processes– Event notification from remote hosts to controller

• Raven– Experiment build environment– Slice management– Resource discovery and selection– Software deployment– Configuration management– Monitoring

PlanetLab, EmuLab, and ORCA

• PlanetLab– The primary large-scale testbed used for experimental overlay– Virtualization tools to efficiently share the global resources

• Emulab– A large-scale network emulator, based on a set of computers– Configuration of various topologies through emulated network

links

• ORCA (Open Resource Control Architecture)– An architecture for on-demand networked computing

infrastructure– A service-oriented resource control plane

Testbed Requirements for Building Service-Based Media Applications

• Experiment Control– Abstraction of the details of underlay– Centralized experiment control

• Experiment Sustainability and Repeatability• Flexible configuration of TB resources and their capabilities

– Real-time experiment measurement• Testbed Management

– Virtualization of computing/networking resources – Ease of testbed management– Real-time resource monitoring

• Resources– High-performance network infrastructure– Specialized hardware for media processing and delivery

FIRST@PC Testbedby the Proposed Testbed Platform

Top-Level Logical Components of the Testbed Platform

• TB management server coordinates multiple RA of the testbed in order to provide resources for experiments.

• TB experiment control server installs, configures, and executes services involved in experiment description.

• PCN (programmable computing/networking) RA (resource aggregate) represents a group of PC-based computing and networking resources in the testbed.

PCN (Programmable Computing/Networking) RA for FIRST@PC Testbed

Realization of Testbed Platform

An Example of Experiment Description using Service Composition Algorithm

• A heuristic service composition algorithm (based on greedy method)– SSSD (single source single destination)

– MSSD (multiple source single destination)

– MSMD (multiple source multiple destination).

J. Liang and K. Nahrstedt, “Service composition for advanced multimedia applications,” SPIE/ACM Multimedia Computing and Networking Conference, MMCN’05, 2005, San Jose, CA, 2005

Experiment Tools (GUI)

• Experiment Control Interface & Experiment Status Viewer– Left: PCN node list & service

properties– Center: a network of services– Right: participating user list

• User Interface– User requirements (for media

sharing) such as End-system capability, desired media source, and media display layout on tiled display

Agents for Experiment Control(Extending SMeet Toolkit)

• Experiment Agent– Being responsible for service configuration,

composition, and monitoring• Node Agent

– Create & close virtual nodes– Start & stop service agents and measurement tools

• Service Agent– Start & stop software tools

• Measurement Agent– Report measurement data to the experiment agent

Real-time Measurement using OML/OMF

• PCN node performance monitoring by OML tool using Sigar API– System memory, CPU load average

– Network interface detection, configuration information and metrics, NIC input/output traffic

• Service performance monitoring by OML tool

Virtualization using

• Create & close virtual nodes in a PCN node

• Add & remove virtual Ethernet devices in a virtual node

• Connect two virtual Ethernet devices attached to different virtual nodes (of a PCN node) using bridge

• Tunneling between different PCN nodes

Testbed Hardware Setup: FIRST@PC MediaX Testbed

An Example of Media Service Composition Experiment for Preliminary Testbed Evaluation

Progress and Plan

• Current Progress– Hardware setup and OMF/OML installation

– Networked tiled display setup

• Plan– In developing OMX (open media experiment for

service composition) toolkit, extending SMeet (smart meeting space) toolkit