BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Talks.cam//talks.cam.ac.uk// X-WR-CALNAME:Talks.cam BEGIN:VEVENT SUMMARY:The Little Engine(s) that could: Scaling Online Social Networks - Josep M. Pujol DTSTART:20100816T100000Z DTEND:20100816T110000Z UID:TALK25807@talks.cam.ac.uk CONTACT:Microsoft Research Cambridge Talks Admins DESCRIPTION:The difficulty of partitioning social graphs has introduced ne w system design challenges for scaling of Online Social Networks (OSNs). V ertical scaling by resorting to full replication can be a costly propositi on. Scaling horizontally by partitioning and distributing data among multi ple servers using\, for e.g.\, DHTs\, can suffer from expensive inter-serv er communication. Such challenges have often led to costly re-architecting efforts for Twitter and Facebook.\n\nWe design\, implement\, and evaluate SPAR\, a Social Partitioning and Replication middle-ware that mediates tr ansparently between the application and the database layer of an OSN. SPAR leverages the underlying social graph structure in order to minimize the required replication overhead for ensuring that users have their neighbors ’ data co-located in the same machine. The gains from thi s are multi-fold: application developers can assume local semantics\, i.e. \, develop as they would for a single machine\; scalability is achieved by adding commodity machines with low memory and network I/O requirements\; and N+K redundancy is achieved at a fraction of the cost.\n\nWe provide a complete system design\, extensive evaluation based on datasets from Twitt er\, Orkut\, and Facebook\, and a working implementation. We show that SPA R incurs minimum overhead\, can help a well-known Twitter clone reach Twit ter’s scale without changing a line of its application lo gic\, and achieves higher throughput than Cassandra\, Facebook†™s DHT based key-value store database.\n LOCATION:Small public lecture room\, Microsoft Research Ltd\, 7 J J Thomso n Avenue (Off Madingley Road)\, Cambridge END:VEVENT END:VCALENDAR