When reading or writing large numbers of records, some
READ_WRITE
operations can take long periods of time. Read
operations generally take longer than write operations for similar size record
sets, and the transaction time of a
READ_WRITE
operation grows approximately linearly as the
number of records grows and the size of the records grows. For this reason,
delta updates are generally faster than baseline updates.
If reading or writing operations cause performance concerns, there are several changes you can make:
Use fast-access drives and RAID with striping on machines that host CAS record stores. This improves disk I/O for better overall record store performance.
Reduce the
generationRetentionTime
setting so that fewer generations of records are stored. For details, see Deleting stale generations of records.Split a crawl into multiple crawls that use multiple Record Store instances.