Concurrent restoring with PostgreSQL

XiVO, as well as XiVO-CC and XiVO-UC, is using Postgresql object-relational database. For easy system maintenance, there’s a periodic cron job doing backups on each installed XiVO. The backup is a operation while restoring is quite time-consuming. Recently we realized that we’re not using the --jobs option of the pg_restore tool, which activates concurrency. Using this option we experienced around 5 times faster data restore on a machine with 8-core CPU. According to the nmon tool, when -j 10 is used on an 8-core machine the hard drive becomes the bottleneck. Have you any experience with the best number of jobs settings?

Jirka