Lazy Replication

Abstract

In this chapter, we have a closer look at lazy protocols. Lazy protocols execute transactions first locally and only send the update information some time after commit. In primary copy approaches, this means that the primary always has the current version of the data item while secondary copies might be stale. In Section 8.1, we present mechanisms that quantify and bound the staleness that can be observed at the secondaries. In Section 8.2, we discuss how lazy primary copy approaches can be made more flexible by allowing the primary copies to be distributed across many replicas. Section 8.3 is dedicated to lazy update anywhere approaches. The particular problem of this protocol category is that transactions executing on different replicas can concurrently update the same data items and all commit. We present mechanisms that allow the detection of such conflicts, and we discuss strategies that resolve these conflicts in order to agree on a final value for the corresponding data items.