1. Caching Consistency and Concurrency Control Contact: Dingshan He he@cs.umn.edu

2. General Problems in Distributed System Object replication to improve performance Multiple clients may simultaneously READ/WRITE the same object Clients may cache either part of an object or the entire object to reduce network traffic Link or system failures

3. New Challenges in Proposed System Mobility –Clients could have high mobility –OSD’s have mediate mobility –Regional Managers are relatively static Connectivity –Disconnection is possible at any place –Clients could have weak connectivity (low bandwidth/long latency)

4. More Challenges Partition of Objects –Objects consists of attributes, metadata, and data –Attributes could be stored at either tier Compound Objects –A compound object refers multiple other objects, either compound or not

5. Levels of Consistency Objects indicate desired consistency level Strong consistency –global consistency at any moment –using server locking Weak consistency –allow temporary inconsistency –using optimistic concurrency control Something in between?!

6. Versions of Objects Optimistic concurrency control generates co- existing multiple versions of objects Version history has to be maintained until final reconciliation Full copy of individual versions consume too much storage Exploiting incremental/differential updates –Existing scheme for text file and database app. –Seek more general support for various objects

7. Distributed Committing Updates of compound object may involve several referenced objects Such updates have to be performed in a transaction way Such updates also have to support versions such that rollback is possible Determine global serializability of distributed updates