Apples and pears comparison
Mirror, mirror on the wall, who has the best TSM backend of all?
A respected EMC blogger and IBM are arguing over who has the best scale-out backend storage repository for Big Blue's TSM backup and archive software: EMC's Isilon or Big Blue's GPFS. IBM says that TSM – Tivoli Storage Manager – products "provide backup, archive, recovery, space management, database and application protection …
COMMENTS
-
Tuesday 12th August 2014 10:43 GMT Anonymous Coward
Quite a long way from apples to apples
The Isilon had 108 spindles while IBM had three times that. A three node Isilon cluster is the smallest it gets and so the smallest stripe set and NL400 are the lowest performance nodes. Isilon is pretty close to plug and play configuration while that GPS configuration will take a bit more work
The GPS numbers are impressive but the solution is probably much more costly and I suspect the end user doesn't need the backup to complete in 20 minutes .
-
-
Tuesday 12th August 2014 19:32 GMT Diskcrash
Either solution will work and will work well but there is a lack in real comparison of the differences. An organisation will choose one over the other not only on performance but costs and other operational considerations.
The Isilon used 108 4TB drives running in a 3 node configuration that took up 12U of rack space.
The IBM used 348 2TB drives running in 6 trays that took up 24U of rack space.
But the size of drives really has no impact on throughput performance where as number of drives does. If the number of drives in the Isilon is tripled the performance is 2.4GB/s compared to IBM's 5.4GB/s which is closer but still gives the advantage to the IBM solution.
But Isilon does have have higher performing nodes that would narrow the gap and the performance difference is likely to be relatively small between the two solutions.
So the one would want consider things like the IBM solution is using InfiniBand where Isilon is using 10GbE. IBM uses its proprietary GPFS protocol while Isilon uses standard protocols like NFS, SMB, and HDFS. Again neither is necessarily better or more right than the other but each may be more suitable for specifics situations.
Personally, the raw single stream performance of GPFS is impressive but not always needed where as the linear scale out and ease of use with Isilon is impressive and probably more useful.
-
Wednesday 13th August 2014 00:25 GMT Nate Amsden
just curious
was looking at the HP StoreOnce 4900, I haven't used this box but the paper specs claim max write speed of 8.5TB/hr (Catalyst performance of 22TB/hr not sure if that is read or write or both?) which google says is about 2.3GB/sec. Max of about 140 drives in 12U of course and 430TB usable with full inline dedupe etc..
The next step up is a big one to the 6500 which goes up to around 17GB/sec (Catalyst is 38GB/sec) which I think is around 560 disks (just dividing the raw capacity by 4TB disks which is what they use) though it seems to take two cabinets for up to 1.7PB of usable(before dedupe etc).
I think the optimal protocol to use with Storeonce is probably VTL over ISCSI. They support NFS/CIFS too but I'd expect performance is less there.
I've never used TSM nor worked at a place that used it but it seems it is a supported software package for StoreOnce.
just had storeonce on my mind based of another article on el reg from earlier today. I use one of their smaller boxes (about to install a 2nd one at another location - not going to use built in replication though don't need it).
-
Thursday 14th August 2014 13:30 GMT Anonymous Coward
Backup, Recovery and objectives
As backup time is usually less critical than recovery a test with various recovery options would have been more valuable but also in considerations of cost. This is one of the most important questions about the overall cost of any backup solution in considerations of time, money and value of the data under the principles of recovery time and recovery point objectives.
Raj