This post is more than 5 years old
134 Posts
0
1753
Hotspares
Is it a good idea to have 600GB Hotspares for 300GB disks?
I know the rebuild time will be greater but generally I am just curious if many have done this and should I create more than 1 HS per 30 disks if I am going to use 600GB HS for 300's
RRR
5.7K Posts
0
April 21st, 2011 07:00
If you only have 300GB drives in production to be protected by hot spares, you should be using 300GB drives. Why use 600’s ? The only reason I can think of is that a new 600 might be cheaper than a 300. Both will do the job and I don’t even see a reason why a 600 should be slower.
1 Per 30 is ok, but consider this: how many drives have you EVER encountered that were broken in the SAME array AT THE SAME TIME ? I’ve had 2 drives at the same time just recently and I’m doing storage now since 2004.
For DMX (years ago) EMC said (I think) 3 HSs for the first 100 and 1 more for every hundred extra drives. Clariion drives don’t break any sooner than DMX drives.
DanJost
190 Posts
1
April 21st, 2011 07:00
As long as the rotational speed is the same, I don't think it'd be any slower. If you have a mix of 300 and 600GB drives in your system you can certainly use 600GB drives as hotspares.
Dan
jps00
392 Posts
0
April 21st, 2011 10:00
Keep in mind, its the 'in-use' capacity of the drive that is rebuilt, not the entire drive. Although, you may have the drive's capacity 100% utilized.
Generally, using hot spares of the highest rpms and largest, in-use capacity is prudent. That avoids the performance hit of getting a slower hot spare, temporarily in the RAID group.
yaribhas
134 Posts
0
April 21st, 2011 10:00
All we have are FC disks and our resident wants just 600GB disks as HS's.
I just checked and the rotational speeds are the same for 300's and 600's. If I understood all of you right I think we are ok here.
What do you guys think about Raid groups with (6+1) configuration. I know it is going to be slower than (4+1) but will save us disk space
And how hard will it be to convert to a thin allocation in the future if we move in raid group direction now?
Thanks!
jimkunysz
259 Posts
1
April 21st, 2011 10:00
a 300Gb can definitely spare to a 600GB. the iops on the disk are identical so no loss in performance. also, it won't take any longer for a 300GB to spare to a 600GB as compared to a 300GB. There is still only a max of 300GB of sizing to write, even though the target is larger.
In regards to having fewer hotspares in a frame because there just doesn't seem to be enough failures to warrant the recommended numbers of spares, I consider that to be a poor decision.
As drive sizes have gotten larger, the rebuild times have also increased in time. It will take longer for a 600GB to spare out versus a 300GB drive thus the risk of a double-disk fault is increased. If we talk SATA, I always do a minimum of Raid6 plus a hotspare per dae (in the NS/Clariion model, VNX may be different).
Plan for the worst expected condition and enjoy the uptime.
Storagesavvy
474 Posts
0
April 21st, 2011 11:00
Rebuild time will be the same as long as the 600GB disk is the same RPM. FLARE only rebuilds the portions of disk that have data (ie: bound LUNs). So if you only used the first 25% (75GB) of the 300GB disks, it will only rebuild 75GB of data to the spare. This also means that if you have 600GB drives that are not full, FLARE can rebuild onto smaller disks (300GB, 450GB, etc) and sometimes will do that even if a larger disk is available.
Storagesavvy
474 Posts
0
April 21st, 2011 11:00
6+1 is a reasonable RG Size and the performance difference will be minimal.
If you want to move to thin later, you will need to create a pool on different disks, then you can LUN migrate the LUNs into the pool. LUN Migration into a Thin LUN in a pool will perform space reclaim and turn the LUN into a Thin LUN automatically.
Storagesavvy
474 Posts
0
April 21st, 2011 11:00
More drives in a single RAID Group will support more IOPS and Bandwidth, however it will slightly increase latency as well. So more drives is faster AND slower, depending on what you care about. A RAID5 group always needs 4 IOs for a write, unless you get a full stripe write, which is more likely with smaller RAID Groups (smaller stripe sizes).
dynamox
1 Rookie
1 Rookie
•
20.4K Posts
0
April 21st, 2011 11:00
6+1 will be faster, more drives support more iops/MB/s. With virtual provisioning you want to add a lot of drives at once because currently you can't restripe new storage in a pool. So if you burn all of your drives on traditional raid groups, it will be tough in the future.
yaribhas
134 Posts
0
April 21st, 2011 11:00
I am probably wrong but does (6+1) config mean that there is going to be extra I/O's used up for writing parity to more disks than (4+1)?
By that theory iIf we wanted good performance than we might be better of going with Raid group(13+1). with enough HS's right?
Why is (4+1) recommended configuration by EMC if these other configurations give better performance.
And HS's are not permanent replacement in clariion correct?