VTL solution??

[djchopps0013]

Ok we currently have a HP VTL storage device we keep critically deemed apps on for 60 versions. We have hit the point where we need to add additional capacity. We have gotten estimates of additional capacity and dedupe licenses for this aplliance and they are rediculous.

So we are considereing two other vendors. Netapp and Data Domain. My question is are there any other TSM shop using these aplliances? Are the working well with TSM?

I know the poll taken points to mostly IBM and EMC. The reason for considering these over others is the dedupe ability and ability to replicate.

[heada]

How often do you need to restore from older than the most recent version? If you do, how long can you wait for the restore?

If you're like me and 95% of your restores are from the most recent version and if it is an older version the restore can wait for up to an hour so or.... consider setting up the VTL as an activedata pool. This lets you have a pool that is no bigger than what the host has (host has 5TB, you only need 5TB in the VTL) and still provide VERY fast restores for 95% of the restore requests.

-Aaron

[djchopps0013]

restore speeds play a small part, but still matter. We currently have critical data on a VTL solution and production data on tape in collocated groups. Management is looking to go tapeless onsite by 4th quarter and totally tape less next year for DR recovery.

How often do you need to restore from older than the most recent version? If you do, how long can you wait for the restore?

If you're like me and 95% of your restores are from the most recent version and if it is an older version the restore can wait for up to an hour so or.... consider setting up the VTL as an activedata pool. This lets you have a pool that is no bigger than what the host has (host has 5TB, you only need 5TB in the VTL) and still provide VERY fast restores for 95% of the restore requests.

-Aaron

[SpiritSoftware]

A lot of the venders are OEM'ing the Falconstor VTL and Dedupe product so I went with the Falconstor product directly and used my own hardware and saved 50% on using the OEM'ed version

There are a lot of issues with the Falconstor Dedupe product and TSM prior to version 5.1 SP1, and as 5.1 SP1 is only been released the vendors will be a few months behind.

Also I did a lot of testing using SAS drives in an HP MSA Array and the performance is almost as good as using a storage array so I will be implenting this solution on our dr site and replicating to it from our Production VTL

Also TSM 5.5 has a new feature for VTL's which will relabel tapes when they become scratch. I highly suggest enabling this option so the VTL and DeDupe appliances know that the data on the tapes is no longer in use

Graeme

[dennis_in_texas]

Please be VERY aware of the issues reported in the following links, before enabling RELABELSCRATCH for your VTL. This issue just about destroyed our V_TAPES, and caused me personally almost two months of non-stop repair work on the damaged tapes (and at least SOME lost data.)

This messed us up SO BAD, that we are reluctant to use RELABELSCRATCH, even with the fix in 5.5.2. I am MANUALLY relabeling v_tapes every day instead. Yes, it is a hassle, but at least it works, and doesn't cause major issues.

BTW...if you are NOT doing some version of relabeling v_tapes, you probably have a BUNCH of tapes that have gone SCRATCH, in your VTL, but the VTL won't let TSM use them. It is the nature of VTL's to do this. We are once again, considering using PHYSICAL tapes, instead of virtual...

http://www-01.ibm.com/support/docvie...id=swg21330242

http://www-01.ibm.com/support/docvie...id=swg21330112

http://www-01.ibm.com/support/docvie...id=swg1IC58862


APAR IC58862 Fix Announcement

http://www-01.ibm.com/support/docvie...id=swg21288273

[SpiritSoftware]

Thanks for the heads up on RELABEL issues, I have only been using it since 5.5.2 and have not seen any issues but I will keep an eye out.

I also understand your pain with restoring the virtual tapes, I recently lost 3 drives in a RAID 6 group which housed a VTL LUN. And I lost 189 Tapes and I am still in the process of recovering the volumes

Also during the past year and half of use with VTL I have not experienced any issues of with the reuse of scratch tapes. Maybe this is a vendor specific issue

[javajockey]

Wow, Reading these post makes me so happy that I did not fall in to the VTL
Trap, despite how hard the vender's where pushing for it. SpiritSoftware, you lost 3 hard drives in a Raid 6? The odds of that happening must be like a million to 1.

[djchopps0013]

Thanks for all the info all. We are currently at the beginning of the process to identify potential candidates for VTL.

javajockey
I dont think its really the technology, it has been out for a while, but rather the way each vender does things a bit differently and how TSM will integrate with them. Like I said we have a HP VTL and are totally happy with performance and stability of the appliance, just not the cost to upgrade. We actually almost got a two fold in perfomrance with this aplliance over what we were seeing with LTO4 drives.

[khearp]

We have implemented Disk Libraries with de-dup equipment with some success. RELABELSCRATCH=YES is a must for VTL. Upgrade to the highest TSM 5 levels you can. Upgrade to the highest devdrvr version, too. There are devdrvr fixes that pertain to preventing library resets from repositioning the virtual tape to BOT. We have lost data issues due to library resets repositioning the virtual tpae to BOT and over-writing the tape labels. This is not a Disk Library issue, it is a devdrvr and TSM issue. The de-dup engine only supports LTO3 technology. Make your VTL tapes LTO3 to avoid TSM server issues. Stub Vtapes only when full to avoid reading/writing data from/to the de-dup equipment. Use disk storage pools as a staging area prior to migrating data to the VTL. Beaware that you can over-subscribe the 4 Fibre Channel paths to the Disk Library. It is better to have more Disk Libraries than a few despite storage capacities. Drive number rules still apply to to Disk Libraries. Do not use a TSM server as a Library Manager with VTLs. The individual TSM servers should manage their own VTLs. You will have performance and administrative issues if you try to maximize storage capacities on the Disk Libraries. What ever your storage capacity is on your Disk Libraries and de-dup equipment do not approach your storage limits.

We do have multiple data centers and replicate data from one data center to another using remote copy over IP or the de-dupe replication services with great success in DR scenarios.

If implemented properly, Disk Libraries can be a good solution.

[eperez507]

If I was you I would look at the ProtecTier product and Data Domain product. They integrate the best with TSM and are the best deduplication technology.

[bkupmstr]

Khearp,
I found your post very interesting, can you please comment more on the type of VTL you are using and if its hardware and software in use.
thanks,
Jim

[magrillo]

I have been using a SEPATON VTL for 5 years now and have been nothing but happy with my choice. Sepaton OEM's it's VTL to HP, but HP only sells capacity upgrades in large chunks. I suggest going directly to Sepaton for a quote to refresh your current VTL with a new one.

[umur]

Has anyone experienced IBM's ProtectTier? I am especially curious about its network replication capabilities. Let's say that I back up 10TB a day, how much data would it send over network to DR site?

[imane2003]

I am in the process of evaluating VTL solution to be integrated with TSM. I receive three offers : IBM ProtecTier, DataDomain DD690 and FalconStor.
What is the best solution ? Waht is the best way to implement it with TSM?

Please share with me your experience in VTL

many thanks

[SpiritSoftware]

Have you considered using the File Device Class, specially with 6.3 just around the corner which will include replication of node data between TSM Servers. This will give you all the same features as a VTL ( Dedupe and Replication ( with 6.3)) for a lot less cost, just need to be aware that the dedupe processing will share TSM resources ie Memory and CPU

[magrillo]

I have used the Sepaton S2100-ES2 for 5 years. It has been very reliable and the backup/restore times in a "real-world" environment are superior to all others (FalconStor, IBM, EMC, etc.). Physical installation took about 2hours Configuring the virtual library, drives, and tapes took about 1 hour. Setup and configuration in TSM and labeling of the tapes took about 1 hour. So, within 4 hours, I was backing up my entire environment to the Sepaton VTL and have never turned back. Sepaton has a newer model with updated hardware and features. We plan to make a purchase of the new model in FY2011 and our current Sepaton VTL will replace the IBM 3584 Library with 8 LTO3 drives as our DR solution. By mid 2011, we plan to be completely tape free !

[GregE]

File Device Class is not going to give you good LANFree options, if you need LANFree.
(interesting info on the 6.3 node replication)

[imane2003]

Hi all,

Thanks for reply

Has anyone experienced IBM's ProtectTier with deduplication?

[GregE]

Search the forum for ProtectTier. There have been some discussions around here that may interest you. You'll then have some people to ask how satisfied they are.

[Jeff_Jeske]

We have an EMC VTL that uses falconstor nodes. We've had it for 4 years (its nearing EOL) and its been problem free its entire life. About 50% of our node backups end up on VTL. The other 50% are split between an IBM 3584 tape library and 200TB of SATA disk deployed as file device class.

We are running TSM 5.5 and have no plans to move to the problem child 6.x anytime soon.

As for File Device Class..... it works okay but relying on windows to manage the disk is a bit scary. All of the data... both pri and copy can be whacked from the desktop. This means file system corruption, virus, or a click happy admin can scram your TSM data with little or no warning.

If I had my way I would do away with FILE DEVC. I'd use tape for all the heavy hitters like database backups and VTL for all the loose files.

[mminman]

Jeff, Do you use disk storage pools for your TSM backups, then migrate to the VTL v_tapes?

[Jeff_Jeske]

^ Yes. We have 12 TSM instances and each instance has a 500GB front end disk pool. We have max size set to 5GB so anything larger skips the diskpool and goes directly to VTL, SATA, or TAPE.

[mminman]

Thanks Jeff .... just what I needed. We are planning to implement a VTL solution and I wasn't sure which direction to go -- continue disk storage pools or to write directly to the VTL. I can see where we'd do both as well, at least initially during migration from physical tape to virtual.

[northeasternat]

I have used the Sepaton S2100-ES2 for 5 years. It has been very reliable and the backup/restore times in a "real-world" environment are superior to all others (FalconStor, IBM, EMC, etc.). Physical installation took about 2hours Configuring the virtual library, drives, and tapes took about 1 hour. Setup and configuration in TSM and labeling of the tapes took about 1 hour. So, within 4 hours, I was backing up my entire environment to the Sepaton VTL and have never turned back. Sepaton has a newer model with updated hardware and features. We plan to make a purchase of the new model in FY2011 and our current Sepaton VTL will replace the IBM 3584 Library with 8 LTO3 drives as our DR solution. By mid 2011, we plan to be completely tape free !

How does deltastore react to TSM reclaims. Being that Sepaton is content aware (reclaims are not backup sets and will likely not provide the relevant content to the scraper) are you experiencing a substaintial amount of data falling into the unsupported pool from the reclaim process? What are your reclaim thresholds set at?

Are you even leveraging data deduplication? if so what are your average dedupe rates?

Do you replicate the VTL to a co-location? if so whats the total processing time look like to ingest, dedupe and replicate your data?

How many SRE nodes are configured within your grid?

Being that Sepaton now makes use of HDS as back-end storage, will you be required to do a forklift upgrade of your existing system?