![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
dr.demento
n00b
- Joined
- Aug 12, 2005
- Messages
- 1
I am working with a high performance raid setup. We have roughly 80TB connected to SAN on 4Gb Fibre channel switches (I recently upgraded us from 2Gb/s switches). An example of the newer hardware we are working with is Infortrend 24Bay unit with 2GB of cache and new Hitachi 1TB ultrastar drives. The older equipment is Satablades by nexsan but they too are maxed out on cache.
The older devices are configured with 256K stripe size. The Infortrend will default to 128K stripe size. But, I got the idea that people here will experience better overall performance with a smaller stripe size say 64K or 32K. Most of the original setup was not configured by me. At the very minimum, I typically select defaults when not sure.
1. The platform is Linux and I have been using IOstat to guage disk load from individual systems. Are there better ways to measure performance?
2. If the writes and reads are smaller, is a smaller stripe size perferable? The stripe size can be set up to at least 1MB on these storage devices.
The older devices are configured with 256K stripe size. The Infortrend will default to 128K stripe size. But, I got the idea that people here will experience better overall performance with a smaller stripe size say 64K or 32K. Most of the original setup was not configured by me. At the very minimum, I typically select defaults when not sure.
1. The platform is Linux and I have been using IOstat to guage disk load from individual systems. Are there better ways to measure performance?
2. If the writes and reads are smaller, is a smaller stripe size perferable? The stripe size can be set up to at least 1MB on these storage devices.