Skip to content
This repository


Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
tree: 79ae8b6eb3
Fetching contributors…


Cannot retrieve contributors at this time

file 36 lines (21 sloc) 0.985 kb

Selena’s Pg FS talk

  • ext3 gives much faster random writes throughput than ext2.
    • Maybe because the seek bundling is faster?
  • But adding a disk to ext2 makes sequential writes much faster.

Only have 4 disks? What should you do?

  • RAID 5 outperforms everything else for random writes.

Make your readahead buffer about 1MB.

  1. Conventional wisdom

“Raid 5 is the worst choice for a database.” Fast for dsequential
writes in our tests.

“Turning off atime is a big performance gain.” Not in our tests,
but 2-3% for free.

“Journaling FSes have worse performance than non-jounraling.”
Turning it off improves perfornance, but there are edges cases and
differenes we could not explain.

“Striping doubles performance.” Performance is better, but nowhere
near double.

Your reader-ahead buffer is not big enough. Make it 8MB.

“JFS was pretty awesome.”

Something went wrong with that request. Please try again.