Pinned topic XIV and V7000 Questions


Performance-wise seems that the XIV Gen3 will be "faster", but the V7000 is more tunable and we can carve LUNs out on a per-workload need to guarantee performance.
One thing I noticed with V7000 (and this probably holds true with DS8K line as well) is that RAID5 seems to be fair more prevalent. We've been used to using RAID-6 more these days (or RAID-DP in the NetApp world). Perhaps RAID5 makes sense in IBM's SAN world with 10K-15K drives due to shorter rebuild times and RAID-6 still with the bigger NL-SAS drives?
In general, anyone have any thoughts on XIV vs. V7000 or had to make this decision before? XIV takes almost zero management other than to create volumes, but I can get a lot more storage and flexibility out of the V7000 and have heard good things about EasyTier.
-
Re: XIV and V7000 Questions
2013-05-21T11:46:36ZThis is the accepted answer. This is the accepted answer.
V7000 uses Raid5 as default for SAS disks. and Raid6 for SATA/NL-SAS disks. Because of Rebuild times, as you mention yourself.
Same goes for DS8xxx.
You can alter it yourself. You are not forced to any raid level.
Since the raids are pretty small, as several smaller raids are added to a pool, you rarely have more than 8-9 disks per raid set. Making rebuild times pretty short.
V7000 is very flexible. Like the XIV, adding a frame, or in the V7000 case, adding a control enclosure costs some. The V7000 can have a controller + 9 shelfs. so when you want to add shelf number 11, it have to be a control enclosure. Up to a max of 4 control enclosures. (40 shelfs with 24 disks= 960 disks per cluster)
The V7000 can control Performance per LUN - Either IOPS or MBps limit can be set per volume. (Note, not both at the same time) You can not do the oposite - garantee performance. Only limit performance. Of course if you limit everything, you kinda garrantee it as well, but then whats the point of storage pooling?
The XIV can do the same thing, I'm sure. The underlying code for the QoS function came out of an Almaden research project called SledRunner, and QoS on XIV is also based on SledRunner.
The Storwize can be setup for zero management if you like. But if you use stuff like Compression and thin provission, you want to lookt at it from time to time to get an idea of how much real space you use / have left.
--
Anders
-
Re: XIV and V7000 Questions
2013-05-27T11:43:42ZThis is the accepted answer. This is the accepted answer.
- AndersLorensen
- 2013-05-21T11:46:36Z
V7000 uses Raid5 as default for SAS disks. and Raid6 for SATA/NL-SAS disks. Because of Rebuild times, as you mention yourself.
Same goes for DS8xxx.
You can alter it yourself. You are not forced to any raid level.
Since the raids are pretty small, as several smaller raids are added to a pool, you rarely have more than 8-9 disks per raid set. Making rebuild times pretty short.
V7000 is very flexible. Like the XIV, adding a frame, or in the V7000 case, adding a control enclosure costs some. The V7000 can have a controller + 9 shelfs. so when you want to add shelf number 11, it have to be a control enclosure. Up to a max of 4 control enclosures. (40 shelfs with 24 disks= 960 disks per cluster)
The V7000 can control Performance per LUN - Either IOPS or MBps limit can be set per volume. (Note, not both at the same time) You can not do the oposite - garantee performance. Only limit performance. Of course if you limit everything, you kinda garrantee it as well, but then whats the point of storage pooling?
The XIV can do the same thing, I'm sure. The underlying code for the QoS function came out of an Almaden research project called SledRunner, and QoS on XIV is also based on SledRunner.
The Storwize can be setup for zero management if you like. But if you use stuff like Compression and thin provission, you want to lookt at it from time to time to get an idea of how much real space you use / have left.
--
Anders
Hi, just like to add that XIV is in general the first whne it comes to supporting new features from OSes, ex.
vmware - VASA, VAAI, VADP, UNMAP (not supported yet, but soon)
MS W2012 - UNMAP, SCVMM, SCOM2012, geocluster
//Roger
-
Re: XIV and V7000 Questions
2018-01-26T19:02:12ZThis is the accepted answer. This is the accepted answer.
Update to a very old thread (because it's been referenced elsewhere).
However, I did want to point out for anyone looking at this, that the current best practice for Storwize is to use DRAID6. It outperforms Traditional R5 (T-RAID5) regardless of workload, and provide higher resiliency and significantly shorter rebuild times. It's the go forward data protection method for Storwize. You can read about DRAID best practices in Barry White's blog and DRAID6 performance in Andrew Martin's blog.