THE RAID 5 thread

nwdtn

Hello DenSter,
This discussion is in 2012,is there any news for RAID5 performance for new NAV and SQL Versions???
Thanks in advance...

kriki

Everything remains as before. RAID 5 is fast for reads but slow for writes. SAN write cache can help for the writes. Separating the LOG file on a RAID 1 or RAID 10 and keeping the DB on RAID 5 helps. Having SSD is better than classic discs.

One change for SQL though: SQL2014 has delayed durability for databases.
In short : it means SQL takes writes in memory, confirms the COMMIT and then it writes it to the disk.
It works like the COMMIT cache on native DB. So if you lose power between the COMMIT and the writing to disk, you lose these COMMITS. You should maximum lose the last few minutes. So you would not lose much data.
And the chance the servers goes down is very low. If these kind of losses are acceptable to get better performance, you can implement it. But I would first try without without using this.

DenSter

I've been out of performance troubleshooting for years, so what I know is outdated quite a bit. I don't think RAID 5 has changed though, so all concerns still apply.

I do think that disk performance has improved a lot over the years, and NAV on SQL Server is a LOT better than it used to be, so I don't know if RAID 5 is still as much of a problem as it used to be.

Find someone who is current on this in your specific configuration and have them analyze your system.

kriki

Basically, they have the same problems as before, but SAN-technology, disk-technology, NAV-on-SQL technology (the commands that NAV sends to SQL) are all improved, so you don't feel the impact so much. Remember also that NAV is not set-based and there is a lot of business logic processing while doing TSQL commands. So it almost never gets to performance limits of the disks.