next up previous
Next: Conclusion Up: Beehive: O(1) Lookup Performance Previous: Related Work


Future Work

This paper has investigated the potential performance benefits of model-driven proactive caching and has shown that it is feasible to use peer-to-peer systems in cooperative low-latency, high-performance environments. Deploying full-blown applications, such as a complete peer-to-peer DNS replacement, on top of this substrate will require substantial further effort. Most notably, security issues need to be addressed before peer-to-peer systems can be deployed widely. At the application level, this involves using some authentication technique, such as DNSSEC [11], to securely delegate name service to nodes in a peer to peer system. At the underlying DHT layer, secure routing techniques [3] are required to limit the impact of malicious nodes on the DHT. Both of these techniques will add additional latencies, which may be offset at the cost of additional bandwidth, storage and load by setting Beehive's target performance to lower values. At the Beehive layer, the proactive replication layer needs to be protected from nodes that misreport the popularity of objects. Since a malicious peer in Beehive can replicate an object, or indirectly cause an object to be replicated, at b nodes that have that malicious node in their routing tables, we expect that one can limit the amount of damage that attackers can cause through misreported object popularities.


next up previous
Next: Conclusion Up: Beehive: O(1) Lookup Performance Previous: Related Work
beehive-l@cs.cornell.edu 2004-02-11