We have some questions about our array. It is a Surestore 12E with six 9.1GB hard drives and six 18.2GB hard drives. We were doing some volume testing yesterday and during the test I was using GlancePlus to watch the status of the system. During that time the Disk Util was at 100% and I got an alert saying that there could be a problem with our disk usage. I am including the arraydisp and was hoping that someone would have the answer to my problem. Also, I would like to know what the disk space used for Redundancy is and how is it used.
Thanks for any help.
--- Disk space usage -------------------- 1. Total physical = 156298 MB * 2. Allocated to LUNs = 110562 MB * 3. Used as Active Hot spare = 17366 MB * 4. Used by non-included disks = 0 MB * 5. Used for Redundancy = 28370 MB * 6. Unallocated (avail for LUNs) = 0 MB * -----------------------------------------
Hi, The "Used for redundancy" is the parity information which is needed to recteate the information in case of a disk failure. This information takes some of the useable disk space. If you add the Allocated and the active hot spare (the useabel space) and compare with the total there is some GB missing. This is the Redundancey.
i believe glance only looks at the most utilized disk on the system. I wouldn't trust it to tell you disk I/O. Try using SAR (you can find all the options on the man pages.
I just tried sar and it doesnt seem to be working for me. After looking at the man pages I checked /var/adm/sa/ and found that there isnt any /sa files.
Glance is much better than sar for this purpose but Glance (or any other host-based tool) is not terribly useful when talking to arrays. All the host knows is that a large amount of I/O is going through what it thinks is one disk. Your AutoRAID is actually spreading the I/O among several disks behind the scenes.
Unfortunately, you have made the classic mistake for AutoRAIDS and allocated all the space for LUN's. For best performance, no more than around 50% of the capacity should be allocated. The arrtay will then run in RAID 1/0 at all times. You at least have active hot spare enabled so that that space is used for additional 1/0 space. In your configuration, about 10% (the most recently used) data is kept in RAID 1/0 while the remainder is RAID5 - with considerably more overlead and thus less performance.
the best answer is "it's magic in there" to quote one CE!!
This value is going to change the more you use the array (the more data is on it that is) The data is duplicated (mirrored) in a R1/0 style, which has good perf for RAID, but poor disk efficiency (doubled data), Raid 5 is a comprimise on disk efficiency versus performance. It is a parity based raid and thus requires controllers to calculate (resulting in slower io) The autoraid decides which data goes in which raid level. It is nice you use the Active Hot Spare, this is used to store in R1/0, if you really need more space you can disactivate HS and use it as LUN capacity.. with the detriment of course of performance.....
One idea I gave my manager is that we could purchase two more 18.2GB hard drives and replace two of the 9.1GB hard drives. That would then give us another +-16GB of space that would remain unallocated. Would that help with our problem?
not much, i would suggest using RAID 1 if you aren't already. but for faster disk, you would need to move to a SAN like EMC, Hitachi (they have fast on board cache) that increases performance. but you are limited with an autoraid as far as performance goes. Especially if you are running Oracle databases etc....
However, for real performance benefits, a correct lvm lun access is the way.. ie access lun 0 thru X, access lun 1 thru Y. that way load is balanced across 2 scsi busses and you don't end up with one bus idle.
