1 / 24

A Novel Approach for Transparent Bandwidth Conservation

A Novel Approach for Transparent Bandwidth Conservation. David Salyers, Aaron Striegel University of Notre Dame Department of Computer Science and Engineering Supported by NSF Grant: CNS03-47392. Introduction. Internet has grown an evolved Simple connectivity → sophisticated applications

Download Presentation

A Novel Approach for Transparent Bandwidth Conservation

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. A Novel Approach for Transparent Bandwidth Conservation David Salyers, Aaron Striegel University of Notre Dame Department of Computer Science and Engineering Supported by NSF Grant: CNS03-47392

  2. Introduction • Internet has grown an evolved • Simple connectivity → sophisticated applications • Point-to-point nature • Increased redundancy • Techniques to reduce redundancy • Active (multicast) • Passive (caching) IFIP Networking 2005

  3. Multicast Techniques • Current Multicast Techniques • IP Multicast • Application Layer Multicast (ALM) • Issues • Require global deployment • Application and/or end-to-end network support required IFIP Networking 2005

  4. Caching Techniques • Types: • Web/media caching • Packet Caching [Santos,Wetherall USENIX 1998] • Benefits: • Simple deployment • Reduces long term redundancy • Drawbacks: • Cannot handle short term redundancy IFIP Networking 2005

  5. Our Solution • Dynamic “Stealth” multicast • Dynamic formation of virtual multicast groups • Multicast only occurs inside a domain • Uses existing multicast within domain (PIM-SSM) • Does not require global multicast support. • Does not require application support. IFIP Networking 2005

  6. Stealth Multicast: Overview IFIP Networking 2005

  7. Key Principles • External Transparency • The end clients and server applications should not be aware that stealth multicast is operating • Limited QoS impact • Stealth multicast should not significantly affect the QoS of the application, specifically the end-to-end delay IFIP Networking 2005

  8. Stealth Multicast Operation • Virtual Group Detection Manager (VGDM) • Key Responsibilities • Signature generation • Background Traffic Analysis • Managing Virtual Groups • Converts to multicast • Creates/Updates physical trees IFIP Networking 2005

  9. Stealth Multicast Operation: VGDM IFIP Networking 2005

  10. Virtual Group Management • Only amenable packets are queued for possible multicast transport. • Non-amenable packets are never queued • Bounded QoS impact. • MHT – Maximum Hold Time • TSW – Time Search Window • PSW – Packet Search Window IFIP Networking 2005

  11. Stealth Multicast Transport • PIM-SSM (Single Source Multicast) • Source = ingress point • Receivers = egress points • Note: • Egress points << # of clients • Source driven changes (VGDM) • Join/Leave operations • Resource management/billing IFIP Networking 2005

  12. State Management • Problem • Preservation of unique client information. • Destination IP/Destination Port • Encapsulation vs. egress storage • Stateless (encapsulation) • Include after L4 (UDP) header • Simple state coherency • Stateful (egress storage) • Unique state stored at egress points • Limited packet overhead IFIP Networking 2005

  13. Scalability Considerations • Queue size (memory) • Needs less than 6MB of queue to handle 1Gbps with a MHT = 5ms • Redundancy detection • Santos, Wetherall [USENIX 1998] • Experimental studies • Itanium 2 – RedHat Linux (user space - libpcap) • Intel IXP IFIP Networking 2005

  14. Simulation Studies - Setup • ns-2 Simulation • Random ISP domain (32 core, 16 edge nodes) • Server Farm (40 source Applications) • Average number of clients per application: 32 • 500ms average inter-arrival time for join/leave events • Server Applications, UDP, exponentially distributed packet rate of 50ms and size of 500bytes IFIP Networking 2005

  15. Types of Transport Compared • Unicast • No multicast transmissions. • Full Stealth • VGDM is at edge node. • Local Stealth • VGDM is at server node. • ALM • Generic version of ALM • Ability to support 5 downstream clients • IP Multicast • Ideal version of IP Multicast IFIP Networking 2005

  16. Client Subscriptions: QoS Delay IFIP Networking 2005

  17. Client Subscriptions: Link BW IFIP Networking 2005

  18. Client Subscriptions: Domain BW IFIP Networking 2005

  19. Conclusions • Limited QoS (delay) impact • Provides multicast benefits combined with ease of deployment of caching • Key Benefits • No application support required • Simple resource management/billing • Directable economic benefit IFIP Networking 2005

  20. Current Work • Stealth multicast prototype • Wireless stealth multicast • Passive Application Layer Multicast (PALM) IFIP Networking 2005

  21. Questions? dsalyers@nd.edu http://www.cse.nd.edu/~dsalyers IFIP Networking 2005

  22. Simulation Studies – Setup (VGDM) IFIP Networking 2005

  23. Maximum Hold Time: QoS Delay IFIP Networking 2005

  24. Other Issues Cont. • Practical Benefit • While well suited for networks that contain a reasonable amount of redundant traffic, it is not envisioned that VGDMs are deployed at all nodes. • Since the VGDM • TCP • Generally TCP traffic is not amenable to stealth multicast. However, we are researching techniques to eliminate this problem. IFIP Networking 2005

More Related