1 / 48

Agent-based End to End Video QoS - Assessment and Prediction

Agent-based End to End Video QoS - Assessment and Prediction. Vinay Chandrasekhar (2), Tyler Johnson(1), Ketan Mayer-Patel (1), Wesley Miaw (1), John Streck (2), Amin Vahdat (3), Mladen Vouk (2) North Carolina Networking Initiative.

ida
Download Presentation

Agent-based End to End Video QoS - Assessment and Prediction

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Agent-based End to End Video QoS - Assessment and Prediction Vinay Chandrasekhar (2), Tyler Johnson(1), Ketan Mayer-Patel (1), Wesley Miaw (1), John Streck (2), Amin Vahdat (3), Mladen Vouk (2) North Carolina Networking Initiative (1) UNC Chapel Hill, (2) North Carolina State University, (3) Duke University I2/09-April-2003

  2. On Cyberinfrastructure(from the Appendix of the Report of the National Science Foundation Blue-Ribbon Advisory Panel on Cyberinfrastructure, Jan 2003) “Cyberinfrastructure makes applications dramatically easier to develop and deploy, thus expanding the feasible scope of applications possible within budget and organizational constraints, and shifting the scientist’s and engineer’s (and educator’s) effort away from information technology development (and manipulation) and concentrating it on scientific and engineering research (and education). Cyberinfrastructure also increases efficiency, quality, and reliability by capturing commonalities among application needs, and facilitates the efficient sharing of equipment and services”. I2/09-April-2003

  3. Motivation:Support of End-User (Work)flows • W-Flow: Order and way in which we do things to achieve results (e.g., plan and operate video sessions, construct services, do research, solve problems, etc.) • Network-end devices (video, sensors, network-based appliance and equipment), computers, storage, networks and associated services and software (“glue”) make sense if they support, with success, the workflows of the intended end-users (a broad base of users). • Case in point “Video on Tap” I2/09-April-2003

  4. What? I2/09-April-2003

  5. Workflow Overlays Internet Service-Specific Plug and Play, e.g., “video on tap” I2/09-April-2003

  6. (Pro-active) Resource (QoS) Provisioning and Management Has Many Aspects • Distributed Computing, Storage, and Communication (capacity, scheduling, availability, persistence, reliability, portability, interoperability, configuration …) • Heterogeneity, Collaboration, Autonomy, Federation, At-Will, On-Demand, Cooperation, Sharing, Independence, Non-intrusiveness, Promptness, Standards-based, Appliance-like, … I2/09-April-2003

  7. (Pro-active) Resource (QoS) Provisioning and Management Has Many Aspects (2) • Security, Credentials, Trustworthiness, Authentication, Access, Authorization, Survivability, Policies, … • Discovery, Life-time Management, “Memory” (state awareness and persistence), Manageability, Reliability, Performance, Scalability, Survivability, Quality of Service, Help, … • Pro-active workflow support (synchronous and asynchronous) – “standardization/industrialization of this part is still in its infancy … etc I2/09-April-2003

  8. (GRID/P2P)? Middleware & Infrastructure Enhances Workflows of End-Users thru Pro-Active, Adequate and Transparent Resource Provisioning Domain Workflows Applications (Middle)ware GRID/P2P Resources OS E/U Inefficiency Communications Hardware I2/09-April-2003

  9. Overview • Video teleconferencing over IP has been an area of considerable interest for North Carolina Networking Initiative (large-scale MPEG2 and H.323 video deployments). • Special interest – “full stack &path” diagnostic and predictive end to end (E2E) QoS management. • Experiences with ViDeNet tools – very positive, but not Open Source software. I2/09-April-2003

  10. Goals & Progress • The goal of the project is to build an Open Source service/agent video quality assessment/prediction tool that can be easily replicated at any university, and whose code base can be improved by the development community over time. • Elements are in place. We are exploring its capabilities and how it may help us construct a “video overlay”. I2/09-April-2003

  11. Where? I2/09-April-2003

  12. NCNI Network (logical)Research Triangle, NC Duke NCSU NCREN NC GigaPop UNC-CH Abilene Cisco MCNC Other Companies operational pending I2/09-April-2003

  13. North Carolina Research and Education Network Elizabeth City Winston Salem Boone Greensboro Rocky Mount RTP Asheville Greenville Fayetteville Cullowhee Charlotte Pembroke RTP RPoP Morehead City NCCU Wilmington Duke • NCREN3 • Increased bandwidth • Increased reliability • Increased resiliency NCSU Qwest MCNC NCSU Centennial Campus UNC-CH I2/09-April-2003

  14. How? I2/09-April-2003

  15. Baseline Video Development Initiative (ViDe) • International consortium of universities promoting the deployment of digital video in higher education. • Working groups in video conferencing and IP telephony, video-on-demand, MPEG4. http://www.vide.net/ ViDeNet • Global mesh of interconnected H.323 zones connecting campuses via the Internet and Internet2. • Provides a testbed for inter-networked video and voice over IP architectures and related technology. • Goals: interoperability, low cost, a global voice and video network. http://www.cavner.org/videnet/ I2/09-April-2003

  16. Other work • H.323 Beacon project (Ohio ITEC) – I2 • Access Grid beacon (IP multicast) - GRID • Commercial tools (e.g., NetIQ) –COTS • Vendor-specific (e.g., Polycom, VCON) • Other I2/09-April-2003

  17. ViDeNet Scout Internet Remote user site B Remote user site A • Are endpoint pairs ready to support this traffic? • Prototype uses Chariot. • End-to-end performance results, not just first to last hop. • Readily deployable. • QoS test configurations possible. • Multi-endpoint network configurations possible. • Remote access. Remote user site C I2/09-April-2003

  18. Example of a mixed voice and video traffic load 64 kbps audio 384 kbps video 64 kbps audio Remote user site A Remote user site B 384 kbps video 64 kbps audio 768 kbps video • Bi-directional. • Multiple streams with heterogeneous requirements. • Possibly asymmetric. I2/09-April-2003

  19. Scouting Advanced Networks10 minute384kbs simulated conference CUDI (Mexico) (unusable connection) SURFNet (Netherlands) (good connection) I2/09-April-2003

  20. Scouting Out ProblemsPublic Health Outreach Project • Remote Health Clinic connected back to Internet2 via xDSL was unusable • Original diagnosis was h.323 problem • ISP refused problem ownership until presented with Scout results I2/09-April-2003

  21. Tulane - LANet SimulationLouisiana Statewide T1 Network • Marginal performance due to widespread T1 architecture. I2/09-April-2003

  22. Example (H.263 stream) Under Load (few % in net losses) No Load I2/09-April-2003

  23. Experiment Send Frame Rate (fps) Bandwidth at sender (kbps) Recv. Frame Rate (fps) Datagram Jitter (microsecs) Number of Frames lost/total frames Number of Datagrams lost/total frames Bandwidth at receiver (kbps) No load 27 276 27 261 (0.2 ms) 0 0 276 Under load 6 64 6 233,013 (0.2 sec) 4.8 per 158.9 15.8 per 299.8 64 Video Simulation & Analysis Tool (VSAT) I2/09-April-2003

  24. Some Metrics • Bandwidth, Jitter, Delay, Loss • Platforms (CPU, memory, OS, stacks, communication protocols, etc.) • Application level (e.g., Frame rate, frame jitter, losses, recovery algorithm metrics, stream types, etc.) I2/09-April-2003

  25. Resource Monitoring • NC State RUM (some other metrics, flows, load, packets, “operational profile” shape) (demo) I2/09-April-2003

  26. More How? I2/09-April-2003

  27. Architecture (1) • Client-based agents, service-based QoS analysis. • Client agents collect data about the application, platform, network performance (currently simulated video stream based probing of real paths, developing new QoS MIB) and estimates video network characteristics then passes that information to analysis services. I2/09-April-2003

  28. Architecture (2) • Data collection and QoS analysis/prediction services – currently only VQM (Video Quality Manager - VQM). In general, any protocol. • Analysis services provide an estimate of the end-user QoS and possibly recommend remedies (app, platform, network) I2/09-April-2003

  29. Top-Level Global Architecture Registries (e.g, UDDI) and Context Gateways e.g., UNC Service Registry Workflow Composer NCSU Services Uses registered services to construct new services and/or workflows. Saves product on a server and registers it. SOAP Service Gateways and Service Agents UNCCH Services Users XML data descriptions WSDL process descriptions XPDL workflow descriptions URL/URI descriptors Etc. Agents and Workflow Tools Duke Services UNC-A Service MCNC Services Workflow Agent “Dials” needed services/workflows and executes/runs through the services/workflow, delivers output to user. GT Service s I2/09-April-2003

  30. Advanced Workflow Support Tools Increase Productivity of End-Users Domain-Specific and Abstract Problem Solving Support & Data Integration Context Mediation 2 A Semantic Mediation User (Matt) 3 Workflow Support GUI + Canonical Workflow Construction & Execution Agents and Agent-based End-User Support Registries Information Wrappers Canonical/Universal Data Transforms & Integration Services/Objects 1 GRID/P2P Middleware & Infrastructure & Resources Computational Resourses Networks Computational Resourses Data Sources Computational Resourses Data Sources Data Sources I2/09-April-2003

  31. Workflow W-Flow: Precedence, Dependencies, Timing, “Memory,” etc., e.g., XPDL Semantics and Context awareness?? Process (e.g., WSDL) Information Flow (e.g., XML) I2/09-April-2003

  32. Context • McCarthy (87, etc.) – formalization of the context idea using objects (operations, values, variables, relationships), e.g., ist(NCSU, Professor(mav)) • Formally represented knowledge is based on a conceptualization: the objects, concepts, and other entities assumed to exist in some area of interest and the relationships that hold among them. Every knowledge base, knowledge-based system, or knowledge-level agent is committed to some conceptualization, explicitly or implicitly. I2/09-April-2003

  33. Ontology • An ontology is an explicit specification of a conceptualization (Gruber93) • An agent commits to an ontology if its observableactions are consistent with the definitions in the ontology • Vocabulary with which queries and assertions are exchanged among agents. • Use of shared vocabulary does not imply sharing of knowledge. I2/09-April-2003

  34. Some Details? I2/09-April-2003

  35. VQM Oracle 5. Oracle returns a quality value 1. Client sends request with characteristics 2. rtp-recast resends video from the library 3. vicdump dumps reconstructed video frames 4. VQM compares original and reconstructed frames Client Oracle System Client rtp-recast vicdump Video & Network Characteristics Quality Library VQM I2/09-April-2003 Wesley Miaw <wesley@cs.unc.edu> Computer Science Department University of North Carolina at Chapel Hill 11 September 2014

  36. Video QoS MIB - Motivation I2/09-April-2003

  37. Pseudo-Video Source Pseudo-Video Sink Internet SNMP MIB SNMP MIB Client Endpoint A Server Endpoint B Java Reporter / Monitor (SNMP MIB Query and Display, Oracle Interface) Monitor Computer VSAT Block Diagram j a v a N.B. Use of IPERF (vbr, cbr) I2/09-April-2003

  38. Video QoS MIB Design Sender Table: Session ID, Test Number, Video Data Bytes Sent, Video Frames Sent, Number of Video Datagrams Sent, Video Frames per Second, Max Encoding Rate, Activity Level, Video Picture Type, Test Time duration Receiver Table: Session ID, Test Number, Video Data Bytes Received, Video Frames Received, Number of Video Datagrams Received, Effective Video Bandwidth, Video Datagram Jitter, Number of Lost Datagrams, Number of Lost Frames, Datagrams Transferred I2/09-April-2003

  39. Screen SnapshotsVSAT GUI I2/09-April-2003

  40. VSAT GUI I2/09-April-2003

  41. Monitor GUI I2/09-April-2003

  42. Bigger Picture - Overlays • Large-scale utility for network services • Web services, multimedia distribution, event notification, application blades, overlays • Challenges • Scalable to 10k’s sites • Adaptive to factor of 1000 spike in load • Fault tolerant: failures are the common case • Dynamically adapt to changes in utility members, client access patterns, network conditions • Under resource constraint: provision for target levels of performance, availability, and data quality I2/09-April-2003

  43. OPUS: An Overlay Peer Utility Service • Dynamically allocate resources to competing services • Based on changing application and network characteristics, SLAs • Create topology based app requirements • Bandwidth, latency, loss rate, cost ($) sensitivity Peering Overlay node App demand I2/09-April-2003

  44. Adaptive Cost, Delay, Resource Constrained Overlays • Model • Nodes self-organize to build efficient data dissemination structure • Pre-specified root of overlay tree • Independent metrics assigned to links: e.g., cost and delay, other resources • For scalability, cannot: • Broadcast • Use centralized information • Perform global locking, global probing I2/09-April-2003

  45. Scalability and Adaptivity • Adaptivity • Must constantly adapt to changing network conditions • Constant probing of peers to determine current best parent • Worst case: requires O(n) state and O(n) probing • E.g., Narada, RON • Scalability • Limit state and probing overhead to O(log n) • E.g., Distributed hash lookup schemes • Is it possible to be both adaptive and scalable? • In large-scale distributed environment I2/09-April-2003

  46. Flexibly Trading Cost for Delay I2/09-April-2003

  47. Concentric Layers of Resources (resources “come” to users, not vice versa) Coverage Cost Access Usability … Delay (ns to ms) Coupling Response Capacity Availability Reliability ... Local Immed. Metro State/National Global I2/09-April-2003

  48. I2/09-April-2003

More Related