Data vs Index Paths


#1

Slide #11 of http://www.slideshare.net/Couchbase/sizing-couchbase-server-cbc2014-clean suggests using different data paths for data and indexes. What does that mean exactly ?

I see the following under the default data folder ( /opt/couchbase/var/lib/couchbase/data )

  • 1 folder per bucket
  • 1 folder called @indexes
  • some other files.

The bucket data and index data appears to be in separate paths already. Should they be on separate disks instead ?

Also, As per http://docs.couchbase.com/admin/admin/Misc/admin-datafiles.html ,
“The default disk path is typically used for development purposes only. If the server node is used for production, configure a different disk path.” . If the couchbase server node has just 1 disk and is mounted on “/”, does it make sense to change the “default” paths ?


#2

The logic behind this is as follows:

Data partition storage, replicas of those partitions, views, and replicas of those views are all b tree structures stored on disk. As this can be io intensive, if there is an option to physically separate the data storage from the view storage (not just logical disk partitioning), this can provide increased performance and reduce the burden on the io system.