What to buy - Dell DR4000 vs Dell MD3600i

[jpeake]

I am evaluating Veeam and AppAssure to replace Backup Exec/tape. Leaning heavily towards Veeam as there were too many issues with AppAssure during my demo period (many server reboots for agent issues, really clunky web GUI etc, feels like an early beta product - but that's another topic)

I am still debating between a Dell DR4000 and a Dell MD3600i as my backup target, and am not sure which would be the better choice. The MD3600i is a little cheaper ($12k vs. $15k, we are Dell Premier) The DR4000 configured with 5.4 TB of disk, but they claim up to 70TB useable due to inline dedupe. The 3600i would use 12 X 3TB drives, so around 22TB after RAID and formatting. That should be enough storage to satisfy our needs for a couple years. I get barely any dedupe out of Veeam now, but have the jobs set to “Local Target” for speed. I could always tweak that and try to squeeze more data onto the disks if needed.

But if the DR4000 really would give me the ratio they claim, or heck even 50TB, that would be awesome. The dell storage engineer I have talked to claims the DR4000 is the greatest thing in the world. They are selling like hot cakes etc etc. But she doesn’t have any white papers or data to show real world dedupe rates.

My concern is if I don't get better than 5:1 dedupe, then the MD is the better buy for sure. And wouldn't the MD be faster as well, since it's not doing dedupe magic? And the $3k I would save gets me nearly half way to the $7k I would need to slap a fully populated MD1200 on the MD3600 in the future to expand it.

But if the DR can actually get close to the 15:1 ratio they claim, or even 10:1, then I think it would be the way to go. And it has some pretty cool replication features, where (in a future budget year) I could place another one offsite, and not have to use Veeam to do the heavy lifting for DR offsite transfers.

Our production storage is a pair of EqualLogic arrays (4100X and 3700X). Total used storage is 12 TB.

I am hoping to hear from folks who might have the DR4000 (or similar inline dedupe storage) and see what kind of results they are getting. I will certainly post detailed results of my setup after I choose storage and am up and running.

Thanks all (another reason Veeam is better than AppAssure - this forum, has been invaluable during my demo period)

[jpeake]

Another thought - for Veeam 6.5:

I plan on using a physical host as the Veeam server, and if it is supported to install Veeam on Windows 2012 in 6.5, I could mount the MD3600i volumes using Windows dedupe. Maybe that combo would give me similar results as a DR4000, while being cheaper and maybe faster? Not sure.

[dellock6]

Talking about the new scenarios made possible with Windows 2012 dedupe, so why not thinking about a server filled with local disks, and avoid both DR4000 and MD3600 directly attached to the server?
I do not know very well Dell servers, but at first look the R720xd could be a good candidate, since it can hold up to 26 disks. Try to do a configuration and see how it costs. you can then install Windows 2012 on it and configure Veeam repository.

Luca.

[chimera]

A timely post, I'm going through the same "considerations" for a client. We're looking at the DR4000 as well, which even better Dell currently have a promotion buy 2 x Dell EqualLogic SAN's get a DR4000 free (client is getting a couple new EQL's anyways, so its a hard offer to refuse) I've deployed the MD3600i before at clients as primary storage (in fact, tier 2 storage alongside EqualLogics again) Personally,... I'd steer clear of the MD3600i. The SAS models are fine, never had a problem, the iSCSI models - endless issues. The EqualLogic is way better, but of course a different price point. In regards to backup storage, from what I've read the DR4000 is pretty good (based on Ocerina which Dell bought) Can't speak based on personal experience with the DR4000 yet though unfortunately. Let me know which route you go down and your experience

[jpeake]

We use EqualLogic arrays for our primary storage. They are excellent, but as you say, different price point so was not an option for my backup storage. I have a MD3600i on delivery, will hope for better luck that you have had with them. Mine will only have a single initiator connecting to it, so hopefully I don't see endless iSCSI issues. Using the 10Gbe controller, and the Veeam server has Intel X540-T2 NIC.

I hope to have it delivered maybe this week. Will certainly post back initial setup results, and long term results after she "breaks in"

[chimera]

Well the fact you have 1. only a single initiator and 2. using decent Intel NIC's and not crappy broadcom, I'm sure you will have no issues. Our main issues were around massively high latency to the SAN - it could only cope with about 3-4 VM's on it, if it got even slightly "thrashed" latency would go through the roof and freeze all VMWare hosts up (or at least, access to them - according to the vmkernel logs, we assumed that the s/w iscsi initiator was so busy with retries it left no resource for mgmt agents). In your case, should not be a problem. Please post bottleneck results reported by Veeam once you get it up and running.

[ryanm]

We are looking at 2 options to make an offsite backup copy job repository to store GFS backups. This would eliminate our tape backup procedure every month and allow us to store 24 monthly backups.

Option 1: Buy a Dell DR4300 with 36TB native storage, and a rated 580TB storage at 15:1 dedupe rate.

Option 2: Expand our existing MD3600i PowerVault with additional trays (12x 4TB) drives as needed to support our data. This would also allow us to use Windows 2012R2 native dedupe post process to save data.

I estimate to have about 400TB of data to store. I'm somewhat concerned with using a proprietary technology, and also not really believing a 15:1 ratio is possible. What are other folks with dedupe appliances getting for a dedupe ratio and would you buy your product again or just use Windows 2012 or other?

Thanks for your feedback.

[Gostev]

Something tells me you did not use search before creating the new topic

Personally, I am a big fan of cheap raw storage and Windows Server 2016 dedupe.
Windows Server 2012 R2 dedupe will not scale even to 10% of your data size.

I have not seen much positive feedback from Dell DR users.

[ryanm]

Well I did do some searching but most of the info is old. I'm guessing not a lot of people were using server 2012 dedupe in the year 2012. So you are saying that Server 2012 won't scale to the size I need, do you know that Server 2016 is better?

[dellock6]

Surely deduplication in 2016 is way better than in 2012:
https://blogs.technet.microsoft.com/fil ... rver-2016/

If you add to the scenario our new per-vm chains available in v9 and not before, now you can think that, with average source data reduction made by Veeam (dedupe + compression + eventually also bitlooker for windows vm), you can have a single backup file smaller than 4TB (the known limit in 2016) storing a VM as large as 8 TB (50% reduction) and be able to dedupe it even more using 2016 deduplication. Few VM's will fall outside of this 8 TB size probably.

Luca

[Gostev]

do you know that Server 2016 is better?

Yes, it is better by an order of magnitude in terms of scalability.

you can have a single backup file smaller than 4TB (the known limit in 2016) storing a VM as large as 8 TB (50% reduction) and be able to dedupe it even more using 2016 deduplication.

Fatal planning flaw detected, Luca remember that you have to disable compression to for dedupe to work, so there is no 50% reduction. Thus, maximum VM size should be under 4TB of used space. Nevertheless, still few VMs fall outside of 4TB used space limit.

[DaveWatkins]

We've just gone through this process and we decided not to buy a hardware dedupe device. The restore speed issues and trying to go to tape just made it too impractical without also buying more disk storage for a hot landing zone

We've instead bought a Dell 3860f with 41 x 4TB drives that we'll run in RAID10 (with a hotspare). Initially we don't plan to use any dedup at all and while we might look at it when 2016 comes along we may very well not try it even then. I'll be creating LUN's such that they don't break the 64TB limit but I have no guarantees some of our VM's won't end up over that 4TB mark and I've already seen 2 instances of backup files not being writable because of fragmentation and _really_ don't want to see it again.