In a large system, you generally have a number of content servers with DAH servers to distribute the queries, and DIH to distribute the indexing. The thread count is particularly important for making sure that your architecture works fast enough.
In mirror mode, make sure your mirror servers are on different physical machines. For optimal performance, DAH must have the same number of threads as the total number of threads used by the child content servers.
If you have a DAH with four child content servers, which have three threads each, the mirror-mode DAH should have 12 threads.
In non-mirror mode, DAH should have the same number of threads as the number of threads in each distinct content server.
A DAH over four non-mirrored engines with three threads each should have three threads.
Handling a large number of index actions can put a lot of stress on the content servers, and reduce your query performance significantly if you are running queries at the same time as indexing.
You can use the following configuration parameters to help improve indexing performance. For more information, refer to the IDOL Server Reference.
[Server] section:
DelayedSync
MaxSyncDelay
[IndexCache] section:
IndexCacheMaxSize
In addition, the IndexCacheFlushPause configuration parameter can improve query performance during indexing.
For particularly large indexing concerns, you might want to set up a 'day server', in which you index the data for a single day. You can then set up a separate system for querying. Overnight, you must export the data into the query server ready for the next day.
|
|