Few weeks ago I had a chance to test VTL functionality on StoreOnce devices in conjunction with IBM i. HP StoreOnce is pretty new product and I know that a lot of people is waiting for some information ‘from the field’. This is my short summary.
Keep in mind that I did these tests for my own purpose, I’m not a journalist who suppose to test all functionality and compare with similar products from the market. I expect some specific functionality driven by variety of configuration which is specific for ‘my’ data center.
I had two StoreOnce devices to do the test, model 5500 and 3520 with VTL replication between them. The GUI interface was exactly the same for both of them.
VTL configuration and management is pretty straight forward, I’m not going to explain how to do it. I will explain few major things which need to be consider during configuration.
When a VTL is created, it must be connected with a FC port on the device. In order to keep FC ports balanced you need to do a proper planning. VTL is always linked with only one FC port. Such relation is made via a SAN zone. There is no additional logical layer which handles communication between an initiator and a VTL.
StoreOnce can emulate LTO5 drives only if it runs TS3500 library emulation.
All write/restore operations were successful. I tried native IBMi commands like SAVLIB,RSTLIB or BRMS commands, for instance SAVLIBBRM. All saves were correctly recorded in BRMS database.
In terms of performance I didn’t notice any major issues. All saves and restore operation were done very fast de-dpulication ratio was high, but remember that I’ve tested the device not in real production environment. I haven’t performed dozen of backups from multiple hosts at a time. I can’t say how it behaves in live production environment.
StoreOnce devices do not provide LUN masking functionality. I’ve been told that it will be delivered in further releases. How it affects the functionality? It might be an issue if you have dozen of Lpars. Because you cannot have a single VTL with hundreds of drives configured and thousands of tapes. With LUN masking feature it would be possible. You could configure drive 1 and assign it only to Lpar A, drive 2 assign to Lpar B only. Such configuration would allow to keep only one BRMS network across multiple Lpars. If you have hundreds of Lpars you have following options:
- Every Lpar gets own VTL with dedicated drives and tapes. Thus, BRMS network is getting completely useless. In the end you have to manage dozen/hundreds of VTLs.
- Spit Lpars to groups. For instance, if you have 100 Lpars. You would need 4 VTLs at least. Because V7R1 can operate with maximum 32 logical drives. If 4 VTLs would be used, then you need 4 independent BRMS networks. Of course, such limitation exists only if you need to use all 32 drives at the same time.
- If you have 100 Lpars, but you know there is no more than 32 concurrent tape operations at a time, you may stay with one VTL only, and cross the fingers that tape allocation/reservation work well.
The last thing which I want to highlight is VTL replication. The replication works on the slot level. So, if a tape being ‘moved’ to a slot and the slot being replicated, the content of the tape is replicated to the other device. There is no a file system logic (pools) as with EMC DataDomain. From one point of view it is very convenient solution, but I found out that in some situations it can be a problematic. In order to explain the issue, I will first try to explain to the concept. The VTL has multiple drives, tapes and slots where a tape should be moved if it is not in a drive. One or many slots can be replicated. But this general configuration for the entire VTL. In my scenario I have tape A00001-A00100 which must be replicated from site A to B. These tapes are writen on site A and occasionally may be read on site B. Also, in the same VTL I need tapes A00200-A00300 which may be written either on site A or B. Regarding where written process happen I would like to replicate the content across. Because all these tapes belong to the same VTL, it is not possible to revert the replication only for specific tapes/slots. It is either all or nothing. So, in order to achieve the goal, I would need multiple VTLs across the sites, one for tapes A000001-A00100 and another one for A00200-A00300, and the same for target VTLs. Actually it is getting very inconvenient for such configuration.
I hope that sharing my experience would be somehow helpful. If you have any valuable comment about this post please leave it below.