DISTRIBUTED MULTILEVEL RECOVERY IN MAIN-MEMORY DATABASES

In this paper we present recovery techniques for distributed main-memo ry databases, specifically for client-server and shared-disk architect ures. We present a recovery scheme for client-server architectures whi ch is based on shipping log records to the server, and two recovery sc hemes for shared-disk architectures-one based on page shipping, and th e other based on broadcasting of the log of updates. The schemes offer different tradeoffs, based on factors such as update rates. Our techn iques are extensions to a distributed-memory setting of a centralized recovery scheme for main-memory databases, which has been implemented in the Dali main-memory database system. Our centralized as well as di stributed-memory recovery schemes have several attractive features-the y support an explicit multi-level recovery abstraction for high concur rency, reduce disk I/O by writing only redo log records to disk during normal processing, and use per-transaction redo and undo logs to redu ce contention on the system log. Further, the techniques use a fuzzy c heckpointing scheme that writes only dirty pages to disk, yet minimall y interferes with normal processing-all but one of our recovery scheme s do not require updaters to even acquire a latch before updating a pa ge. Our log shipping/broadcasting schemes also support concurrent upda tes to the same page at different sites.