According to this link, it's possible to send the datastore buckets to different disk volumes to take advantage of different RAID configs.
http://wiki.splunk.com/Deploy:BucketRotationAndRetention
Which bucket is associated with realtime searches? Is it always "Hot" or is the data cached somewhere else?
We have an app that is dependent on several realtime searches and we're trying to optimize performance.
thanks
Hi,
All I've read indicates that data hot off the indexing pipeline is put directly into the hot buckets.
yes it's possible to split an index into a location for hot-warm and cold storage. We actually do this, storing our hot/warm buckets on local drives and cold buckets on a network volume. After we did this we did see an increment in performance in terms of events per seconds read during searching and events indexed per second.
You can configure the different locations in indexes.conf like so:
[main]
homePath=/local/myindex/db
coldPath=/remote/myindex/colddb
thawedPath=/remote/myindex/thawed
Making the switch over on a live index is another issue though 😛
Good luck!