Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

This question already has an answer here:

Recently, we bought a "HP-DL380 G6 Server" with 6 146GB (SCSI)HDD for our colleague course management application and website with 10000 daily visitors. we want to choose the best RAID level. how can we choose the right RAID level ? what is the best RAID level for our application ?

share|improve this question

migrated from stackoverflow.com Mar 13 '10 at 14:01

This question came from our site for professional and enthusiast programmers.

marked as duplicate by Zoredache, Sven, Ward, cole, Falcon Momot Oct 12 '13 at 3:37

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

1 Answer 1

For anything with a database on it, you will want separate volumes for the database logs and the data. Set the system up like this:

  • Two disks in a RAID-1. Partition this between a (say) 50GB system area and a 90GB 'logs' volume (these sizes can vary depending on your requirements). Having the system and logs separate makes it easier to upgrade or reconfigure the server later.

  • Four disks in a RAID-10. Put the data area on this drive.

Configure the NTFS volumes for the data and log areas with large (64k) block sizes as this will give slight efficiencies in databaase performance. Set up a large-ish (64k-256k) stripe size for the RAID-10 volume.

On a really busy transaction processing application (with a large random access workload) smaller stripe sizes may give you better efficiency in cache utilisation on the RAID controller as accesses for single block disk reads and writes from the DBMS don't drag so much irrelevant collateral data into the cache. For a more general workload larger stripe sizes will give you better sequential disk access performance.

This layout will give you better database performance than a single volume, as sharing logs with busy data drives creates disproportionate performance bottlenecks.

share|improve this answer