WL#603: Optimized DiskIO handling for new keycache

Status: Un-Assigned

There are some tricks which can improve IO performance of Keycache for some
workloads
1) Index pages should be alligned index page size as this can be used to match
them to File system block size and raid chunk size.

In 5.0 we should add support to specify index block size per index.

2) IO for larger blocks shall be done the same as the block size, instead of key
cache block size. This removes the need for file system to read the block before
writing it back, for large blocks.

2) Is already done with the new key cache in 4.1