Tegile hybrid storage

[depps]

Ladies and Gents,

Anyone here using Tegile storage?

We're evaluating Nimble, Tegile and Tintri.

Tegile seem to offer the most functionality and are priced very competitively. Nice pieces of kit with dedupe and enterprise MLC SSDs too.

Hoping someone here has some real world experience with them

The biggest thing for me is that it sounds like Nimble are soon going to be integrating their hardware snapshots ala. HP and Netapp whereas Tegile being the smaller play will be waiting sometime for that interaction.

Its funny that Veeam is defining our storage purchase, just goes to show how good this software is!

Dan

[rcommins]

Hi Dan -

Disclosure: I work with Tegile.

If you want a great reference point, check out the webinar we recently did with Veeam and a joint customer, Mizuno. Tegile really helped move the needle for them. Here are a couple links:

Webinar Recording: http://bit.ly/1ljPuSq
Written Case Study: http://bit.ly/1p0H78I

FYI - Tegile has some nice snapshot integration done with VMware and Microsoft already as well.

Please let me know if there is anything else I can help with.

I hope this helps,
Rob

[SLogik]

Hello depps,
We are using Tegile for our production Hyper-V cluster using their hardware VSS provider with Veeam. Very happy with the performance for the VMs as well as dedupe and compression on the backups. Performance far exceeds the NetApps we were using. It took a few weeks to find the ideal settings between the Tegile and Veeam to get the best dedupe between fulls; we've been great since then. Do you have any specific questions I could try to answer?

[veremin]

Do you have any specific questions I could try to answer?

So, you're using two Tegile devices - one as production datastore, the other (deduplication unit) as a target for backup jobs, right? What particular settings allowed you to achieve decent deduplication rates? The information about both job (compression, deduplication) and device settings might be helpful for future readers. Thanks.

[depps]

Our plan is to have two arrays, backing up locally and remotely to the arrays so a single array for both production and backup data.

If the production array or whole site goes offline we have backups on the DR array at the DR site.

We're ideally going to keep SQL snaps at 15 minute intervals for 2 weeks then weekly snaps for 4 weeks then monthlies for 12 or 24 months then yearlies indefinitely.

We will also have backup copies on two QNAP's at each site possibly tape or with a cloud provider (iLand) via Veeam 8 cloud gateway.

I guess the questions are

Since you have no hardware integration with Veeam so will still suffer VM stun issues - any roadmap for Veeam integration? We have high SQL change rates and VMWare shapshots of our SQL environment can take too long (due to change / snap commit) causing some mission critical apps to crash. Hardware integration seems to be the solution however SAN snapshots arent often application aware.

Does Tegile offer application aware functionality, if so does the storage have to be direct attached to the VM? Veeam cant backup VM's that have direct attached storage.

What sort of dedupe ratio do you see with Veeam backups on Tegile?

Do you have a support/community forum like Veeam and Nimble?

It appears not many people on these forums are using Tegile?

[SLogik]

So, you're using two Tegile devices - one as production datastore, the other (deduplication unit) as a target for backup jobs, right? What particular settings allowed you to achieve decent deduplication rates? The information about both job (compression, deduplication) and device settings might be helpful for future readers. Thanks.

Hello v.Eremin,
Only one Tegile device, an HA2100. It's active/active controllers, so one controller is serving up storage via Fiber Channel to our Hyper-V cluster, the other controller is via CIFS shares providing a target for the backup jobs.

To provide some additional background, we have several large VMs (2 TB) that we have the need to keep on disk for at least 6 months. The goal was to get dedupe between the large weekly Veeam fulls for these VMs. After moving Veeam jobs from CIFS shares on our NetApp to the Tegile, our bottleneck shifted from target to source (Our VMs were still on the NetApp) and backup job times were about 10% faster. Space savings between backup jobs with the default settings was ok, but not as high as I had hoped. After some trial and error, we almost doubled our space savings with the following settings:

Veeam:
"Align backup file data blocks" enabled
"Deduplication" disabled
Compression set to "none"
Jobs are incrementals with a weekly active full backup

Tegile:
Deduplication on, for compression either gzip or lz4. (We used gzip initially and it gets better compression, however it is CPU intensive. with the number of simultaneous jobs we were running we were at 80% CPU, so we switched to lz4 which dropped it to no higher than 40%. The HA2100 we have is the entry model, higher models have additional CPU resources. I'd recommend GZIP for backups on any model above the HA2100).

With dedupe and compression, we have 56 TB of backups taking up 7 TB of space on the Tegile. While I've seen better dedupe with other products, I haven't seen better performance with it. Whether you are doing a restore from 6 months ago or the previous night, the speed is the same, no performance penalty for dedupe and compression. After migrating our VMs over to the 2nd Tegile controller from the NetApp, backups of those VMs sped up another 30%. Backups are almost (but not quite) twice as fast.

A few things to consider:
1) With dedupe and compression off in Veeam and taking place completely on the Tegile, if you archive to tape or cloud your backup files will be larger
2) In our environment, I was surprised that there was no difference in backup times with Veeam dedupe and compression off, and with it on. If the backup takes 4 hours with dedupe/compression off, it still takes 4 hours with it on, even though on some jobs it's sending 75% less data. YMMV.

Next year we plan on adding a second larger Tegile to take over the production VMs currently on the HA2100; the 2100 will be moved to our CoLo for replication. Veeam is currently replicating VMs to our CoLo with storage on a NetApp.

[SLogik]

I guess the questions are

Since you have no hardware integration with Veeam so will still suffer VM stun issues - any roadmap for Veeam integration? We have high SQL change rates and VMWare shapshots of our SQL environment can take too long (due to change / snap commit) causing some mission critical apps to crash. Hardware integration seems to be the solution however SAN snapshots arent often application aware.

Does Tegile offer application aware functionality, if so does the storage have to be direct attached to the VM? Veeam cant backup VM's that have direct attached storage.

What sort of dedupe ratio do you see with Veeam backups on Tegile?

Do you have a support/community forum like Veeam and Nimble?

It appears not many people on these forums are using Tegile?

Hello depps,
I'm afraid my knowledge is primarily in Hyper-V, so I can only answer your questions from that perspective. I'd be surprised if someone from Tegile couldn't answer your questions as they relate to VMware.

Tegile does have a hardware VSS provider. When installed on a server and configured on the Zebi itself, snapshots are application consistent, whether I take them from Veeam or on the Zebi itself. You just select if you want the snap to be Quiesced or not. Our storage is fiber channel attached and presented as CSVs to the cluster nodes. Are you looking for the integration Veeam has with HP storage now and NetApp upcoming?

I see anywhere from a 3:1 to 8:1 ratio with dedupe and compression depending on the settings I use in Veeam.

There is a support website with best practice guides, webinars and documentation. Support is where they really win you over. Here is my experience, every time: I call their support number. Within one minute, someone answers the phone. That person works with you on your issue until it's resolved, and they resolve it quickly and efficiently.

[veremin]

Only one Tegile device, an HA2100. It's active/active controllers, so one controller is serving up storage via Fiber Channel to our Hyper-V cluster, the other controller is via CIFS shares providing a target for the backup jobs.

But even with a redundant controller you're still storing both production and backup data on one device. Aren't you afraid that if something happens with the underlying storage, you would lose both?

Whether you are doing a restore from 6 months ago or the previous night, the speed is the same, no performance penalty for dedupe and compression.

That's interesting. Could you elaborate on that portion? You mean there is no penalty for rehydrating already deduplicated data?

Thanks.

[SLogik]

But even with a redundant controller you're still storing both production and backup data on one device. Aren't you afraid that if something happens with the underlying storage, you would lose both?

VMs are replicated via Veeam to our CoLo and Veeam backups are archived offsite. It is still a concern until we have a 2nd Tegile at the CoLo that we can replicate snapshots (from an RPO standpoint).

That's interesting. Could you elaborate on that portion? You mean there is no penalty for rehydrating already deduplicated data?

Thanks.

Correct. No penalty for compression or dedupe (or thin provisioning). Our production VMs are deduped, compressed and thin provisioned. Dedupe is not a post process, it is inline.

To make a high level comparison without getting into RAID groups, spares and the like to our existing NetApp - including controllers, the aggregate our VMs were on is taking up 14U of space with 15k SAS drives. The Tegile aggregate, including controllers, is taking up 3U of space, with 7.2k NL SAS drives and 400 Gig of flash. With the inline dedupe and compression, the Tegile holds twice as much data and produces 4 times the IOPS. This is what we have seen in our environment and with our specific configuration, so again YMMV. During testing, we spun up several VMs and physically pulled the controller serving the VMs out of the chassis; There was no noticeable affect on the running VMs, they failed over flawlessly.

BTW, for depps, I checked the manual and see configuration options for Application consistent snaps with VMware; on the backend the configuration looks similar to what we performed for Hyper-V. I would suggest giving them a call if you haven't already, they should be able to tell you if they can do what you need them to or not. Is there anything else I can try to answer for you?

[veremin]

VMs are replicated via Veeam to our CoLo and Veeam backups are archived offsite. It is still a concern until we have a 2nd Tegile at the CoLo that we can replicate snapshots (from an RPO standpoint).

Ok, you've taken care already about that.

Is there anything else I can try to answer for you?

Nope, you seem to have answered all the questions I had. Thank you.

[meilicke]

I have tested all three arrays in our datacenter, but not around Veeam and backup, just as primary storage. Our use case was for production VMware storage. If you have any questions about them, I'd be happy to tell you what I can. I was happy with all three.