1 / 30

CS 268: Project Suggestions

CS 268: Project Suggestions. Scott Shenker and Ion Stoica (Fall, 2010). Overview. Will present 19 project suggestions Legend: based on how well-defined projects are, not necessary how difficult they are Well-defined projects Less-defined project

tamanna
Télécharger la présentation

CS 268: Project Suggestions

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. CS 268: Project Suggestions Scott Shenker and Ion Stoica (Fall, 2010)

  2. Overview • Will present 19 project suggestions • Legend: based on how well-defined projects are, not necessary how difficult they are • Well-defined projects • Less-defined project • You need to define project • Need to send us a one page proposal by Sep. 22 • Feel free to talk with us beforehand! (in fact, we insist!)

  3. Outline • Wireless • Software-Defined Networking • Congestion Control • Security • Economics • Datacenters • Network architectures

  4. Wireless

  5. Rapidly fluctuating networks • Design a routing algorithm that is targeted at networks where the link quality fluctuates rapidly • Fast compared to global recomputation time • Slow compared to packet transit times • Graph relatively stable (i.e., not about mobile nodes) • Tradeoff between efficiency and performance • Flooding is a baseline: inefficient, but packets arrive! • Are there theoretical limits on performance? • Is there a literature on this problem?

  6. Network Coding • How well does it work, and why? • Revisit SIGCOMM 2006 paper on topic • Performance w/ TCP and w/o TCP quite different • Look at different load patterns, identify what conditions make network coding work well

  7. Software-Defined Networking

  8. NOX at home • Build a simple system that people can use to manage their home network: • Implicit identification • Easy access control • Security measures (identify bots, limit spam, etc.) • Internal debugging • External debugging • Etc. • Not rocket science, but could be widely used!

  9. Analysis of SDN • Can one theoretically characterize SDN: • Complexity? • Reliability? • Performance? • Compare to current distributed approaches…

  10. Congestion Control

  11. Dueling Diatribes • Respond to Bob Briscoe • In public

  12. Blending Paradigms • Can one combine the “fairness” religion with the “pay-for-congestion” religion? • One can consider two timescales: • Fairness on short timescales • Payment on longer timescales (for “share”) • Are FQ and Kelly just extremes along a sensible continuum?

  13. Datacenter Implosion • Is the problem real? • Come up with another solution to the problem

  14. Decongesting the Datacenter • Decongestion: great idea whose time never came • Datacenters: new context where deployment might be possible • Is this a good marriage of opportunity and answer?

  15. Security

  16. Phinishing Phishing • Simple method to address phishing: • Ask people who they think they are talking to! • Build a prototype system for this • Limit number of user interruptions • Identify what new global services are needed

  17. Living with Secure Hypervisors • Assume that every host has a secure hypervisor • What does that mean for security? • Not-a-bot • VDC • What else can we do? • How would this change the world?

  18. Living in a Google World • The existence of large-scale infrastructures like Google allows us to assume that there are Internet-scale systems that can deploy new services. • How does this change: • Security? • Deployment of new architectures? • Test case: deployment of flat names

  19. Economics

  20. Network Neutrality • Is the Google-Verizon pact a good thing? • What does this mean for the future of the Internet? • Can you back this up with a model?

  21. Datacenters

  22. Comparison of Datacenter Routing Topologies • Many datacenter routing topologies proposed so far • Portland, VL2, BCube, … • Compare these proposals in terms of • Scale and incremental scaling • Number of ports • Wiring • … • Questions • Is there one answer? • If not, when should we use a topology and when should we use another?

  23. TCP for Datacenters • TCP not adequate for datacenter environments • Very low latency, high link capacity, low loss rates • Optimize TCP or invent another flow control protocol to • Reduce impact of packet loss • Optimize flow-start • … • You can assume router/switch support • See DCTCP paper at SIGCOMM’10 for related work

  24. Cross-Layer Optimization for Datacenters • Widely different workloads • Latency-bounded requests (e.g., search queries) • Large file transfers (e.g., data replication) • How could you optimize the transport protocol if you knew the type of traffic? • E.g., avoid slow-start for short latency bounded-requests • What would be the mechanism to pass application “hints” to transport layer? • [Optional] How would you “protect” the transport layer against misbehaving/malicious applications?

  25. QoS in Datacenters • QoS has mainly failed in the Internet • Is there a case for QoS in datacenters? • If yes, what is the service model? • Reservation? • (Weighted) Fair sharing? • Differentiated service? • What are the challenges?

  26. Small-Scale Multicast • Data replication: common workload in datacenters • E.g., GFS, HDFS, a block is replicated two or more times • Optimize this communication pattern • Design a multicast solution for a small number of receivers, e.g., no more than 10 • Challenges: • Which layer? • Flow control • Reliability

  27. Network architectures

  28. Burst Switching • Two main communication models • Datagrams: each packet is individually switched (routed) • Circuits: a circuit is set-up and all packets are forwarded • Hybrid model: burst switching • First packet describes how many packets are in a burst • Router decides whether to forward all packets in the burst or none of them • Research • Design a burst switching protocol and study its trade-offs

  29. Caching Everywhere • Assume • There is caching in every router and switch • 90% of traffic will be video by 2013 (CISCO report) • Questions • What is the impact on backbone traffic? • What is the impact on ISP policies? • Study • Assume different video access patterns • See us for possible traces

  30. Next Step • You can either choose one of the projects we discussed during this lecture, or come up with your own • Pick your partner, and submit a one page proposal by September 22. The proposal needs to contain: • The problem you are solving • Your plan of attack with milestones and dates • Any special resources you may need

More Related