Nomadic Information Access课件.ppt

Chord: Internet-Scale Hash Tables N32 N10 N5 N20 N110 N99 N80 N60 Circular ID Space Chord maps ID’s to “successor” Nodes and blocks have 160-bit ID’s Successor: Node with next highest ID Stores block IDs 21..32 Block IDs 33..60 精品文档 Basic lookup N32 N90 N105 N60 N10 N120 K80 “Where is key 80?” “N90 has K80” Successor pointer 精品文档 “Finger table” allows log(N)-time lookups N80 ? ? 1/8 1/16 1/32 1/64 1/128 finger[k] points to successor (n + 2k) log(n) fingers in all R = log(n) immediate Successors for robustness Stabilization methods for concurrency 精品文档 RON/CFS Deployment ~20 RON nodes running CFS servers in 4 continents Many different Autonomous Systems KAIST, Korea Venezuela MSR, UK 精品文档 Robustness Against Failures Failed Lookups (Fraction) Failed Nodes (Fraction) 1000 CFS servers Average of 5 runs Run before stabilization All failures due to replica failing 50% of nodes disappear but only less than 1.6% of lookups fail 精品文档 INS: Nomadic Resource Discovery [service = printer] [building = schafer [room = *]] Intentional name Late binding: integrate resolution and message routing Lookup schafer-printer@ Resolver self-configuration Intentional name resolvers form an overlay network Uses Chord as underlying hash lookup for scalability [service = printer] [building = schafer [room = 405]] Intentional name 精品文档 精品文档 SFSRO runs on the client. Expects a block DB, indexed by block identifiers. Originally single DB. Dhash provides a distributed block DB. Chord helps Dhash find blocks. Will talk briefly about SFSRO FS format. Then talk about Chord. Most of talk about Dhash. Just need to make progress, and not overshoot. Will talk about initialization later. And robustness. Now, how about speed? Small tables, but multi-hop lookup. Table entries: IP address and Chord ID. Navigate in ID space, route queries closer to successor. Log(n) tables, log(n) hops. Route to a document between ? and ? … 1000 servers. Avg/min/max of 5 experiments. 1000 looku