Kris, Karthik, Ansley, Sean, Jeremy Dick, David K, Frans, Hari DHT Routing Kris, Karthik, Ansley, Sean, Jeremy Dick, David K, Frans, Hari
Churn Dealing w/ faults & failures App interface Proximity & topology Today’s Internet Tomorrow’s Internet Security
Churn Given: Evaluate performance in terms of: Churn model and rate Bandwidth consumption Application workload Evaluate performance in terms of: Lookup latency Insert latency Q: How do different techniques & protocols compare along these axes? What works best under each set of conditions? Example Structured DHT routing v. unstructured Do “ultra peers” make sense? One-hop routing v. log N routing + aggressive caching
Dealing with faults & failures Network partitions & failures Healing strategies Coping with non-transitive Internet routing Large discrepancies in node-to-node performance Mechanisms for choosing alternate paths when failures are detected?
Application Interface Generality How independent can applications be of DHT routing, and get acceptable performance? Should we study application-specific DHT routing? Should we shoot for one big DHT? Iterative v. recursive resolution Move the data through the overlay or not? Evolving the API from IPTPS-II k of N API Degree of redundancy in routing paths Traffic priorities within DHT
Proximity and topology Picking good Internet paths Picking good placement points Goal: Good performance & failure resilience What DHT techniques can be used? Scalable probing, synthetic coordinates,… What can the IP layer tell us (e.g., mapping info -- IP-layer topology, past failure statistics info, link bandwidths, etc.)
Interacting with today’s Internet Coping with network reality Coping with ISPs blocking/limiting P2P Coping with middleboxes, NATs, firewalls… Helping them cope Traffic engineering Congestion control, traffic prioritization within DHT P2P proxy caching Many coexisting, potentially competing
Internet architecture Q: Can DHTs help evolve the Internet architecture? Location-independent naming Use DHTs to enable informed application routing (e.g., weather service, outage avoidance, etc.) Stitching the Internet back together Across NATs, firewalls, middleboxes Do DHTs need policy routing?
Security See security group