Set ATA and FC Disks in different bus is good for performance.But you should unbind the luns and destroy the raidgroups on your ATA disks.Change the Enclosure ID.Then connect to the other BE port.
NST or USM both works.USM is the new version of NST.
I removed luns raid groups and hot spares. I had to rearrange the enclosures to keep it simple for cabling, it's my first time to do this (thats what she said).
The only thing I did not like is there were more FC encl than ata. So on ea1 i had to put fc encl first then ata on the rest.
I hope that the encl id is set correct, on ea0 i started with 1-3 then on ea1 1-4. i left the dpe alone at 0-0.
First i powered on the fab switches and made sure they were booted up, then all dae's one at a time while waiting 5 min, now i am
going to power on the dpe and wait for it to boot 25 min. not sure what it is going to look like in nst.
am i missing any thing any performance hints for setting up raid groups and luns for disk to disk backups
dang i am tired have been at it to long so if this post doesn't make sense forgive me but need to get backups going againn
I set the FC Enclosures on b0 and the drives as 4+1 rg3 created lun's also as rg3, on the ata's set them as rg10 4 drives and lun same. This is our backup san untill we can upgrade the cx4. I have it setup to spool no retention so I can destroy rg's and learn. on the fc it did full backups fast no soft scsi errors, but on the ata i got a ton of soft scsi errors. there are 4 fc enclosures and 3 ata so on b1 i had to make the first encl on b1 fc and ata on others. shouls i try rg6 or rg3 on ata? i thought about going across enclosures and also meta luns. looking at performance it is getting hammered, spb is slightly higher than a. when creating i just did auto on sp's and it looks close on balance. the cx500 is up on latest flare 26 and drive fw. still trying to get a feel of looking at analyzer performance to adjust as needed. it usually doesnt have the load that i put on it cause not only doing full backup's but also auxcopying to lto4 at main and dr sites.
Flare 26's Raid5 using a modified algotithm. So you can use Raid5 instead of Raid3.
Ok , I will need to google modified algotithm. would that be alignment offset changing to block size,Or you talking about raid5 the way parity works on each disk instead of raid3 one drive for parity.
The fiber channel drives are raid3 that seem to be doing fine, it is the slower ata enclosures that are giving so many soft scsi errors.
So on the ata hd's setup for raid5 and use more drives to spread the work load?
FrankLiu
1 Rookie
•
45 Posts
993
0
Posted November 26th, 2010 04:00
Oops, the word is 'algorithm'.
Modify include combine data in cache to make full strip write;change parity location every 8 strips and so on.
That makes Raid5 's sequential write performance like Raid3 ,but Raid5's paraller i/o is better than raid3.