Aggregation difficulties content tailored to each user filter item with privacy checks for each user impossible to aggregate and filter when content is crated resolve data dependencies and privacy check each time content is viewed pull vs push social graph? extreme read demands on graph data store Memcache rapid deployment data mapping cache invalidation client code that is deployed frequently created abstractions for graphcs r/w objects (nodes) associations (edges) direct access to MySQL deprecated for graph data tyles Tao service implements objects and association model motivation encapsulating failures in the PHP API access graph easily from non-PHP serivces problems with lookaside cache architecture Inefficient edge lists KV cache is not good semantic fit for lists of edges queries must fetch the entire edge list list support would help make changes to single edge causes entire list to be reloaded requires coordination of incremental updates to cached lists Distributed control logic L-A $ control logic is run on clients clients don’t communicate with each other increases the number of failure modes difficult to avoid thundering herds Nishtala et al.

Requirements real-time aggregate dispersed data access hot set scale refs [1,2,5,6,12,14,34,36] Front-end cluster read heavy workload (100:1 R/W) wide fanout handle failures 10 Mops/s Q: what is a wide fanout Multiple FE clusters single geo region control data replication data consistency 100 Mops/s Multiple regions muliple geo regions storage replication data consistency 1 Bops/s Pre-memcached High fanout data dependency graph for a small user request Look-aside cache why deletes over set