Drobo Users?

[btkrausen]

I have the same setup in 3 different remote offices - all have a B800i directly connected to my proxy server via iSCSI, each has 4 x Western Digital RE4 2TB drives. Currently the highest throughput I've seen is 80 MB/s with MTU set at 9000 on both sides. Although I'm pretty satisfied with that speed, can I expect it to get better?

Also curious to see what speeds you guys are getting with your Drobo units.

Thanks in advance.
Bryan

[Gostev]

Although I'm pretty satisfied with that speed, can I expect it to get better?

Not with B800i... I used one - you are lucky to get the current speed! It uses Marvell CPU, which is pretty weak, so this storage cannot even saturate 1Gb Ethernet. I always recommend looking at faster x86 processor based NAS, using Synology DS412+ myself.

[btkrausen]

Thanks Gostev. These are used for 3 remote offices which have only 6 VMs at each location so a "SAN-like" speed isn't required per say, I was just curious to see what other people were getting. If I continue to get 80 MB/s than I'll be a happy admin for now. When I hit close to 80 MB/s the bottleneck seems to move from the target to the source at that point (the proxies are running network mode).

We weren't even using Veeam at these locations yet (although we purchased the licensing to do so) but these locations seem to be prone to natural disasters - New Jersey(storms) and Colorado Springs(fires). The ability for me to spin the Drobo down and somebody take the appliance offsite is nice. Plus, we might start using DFSR or the Drobo copy feature to replicate the backups offsite. Veeam really came in handy during Hurricane Sandy in moving a 500GB file server from Princeton, NJ to one of our main datacenters. Accounting employees praised our ability to have their file server up and running when the power to their office was out for over a week.

[Dave-Departed]

I have a very similar setup, the difference being that our B800i is connected straight into the SAN network switches instead of the proxy server, and I see reports of around 300mpbs regularly.

Are you using double parity on your Drobos? Initially, we had this configured, and saw horrendous speeds/reaction times when working with VMs... Drobo actually state that you should only use single parity when configuring a Drobo for use with a virtual environment.

Thanks

Dave

[foggy]

and I see reports of around 300mpbs regularly.

Are these probably Megabits/s (giving us < 40 MB/s) or estimate of an incremental job run?

[btkrausen]

I have a very similar setup, the difference being that our B800i is connected straight into the SAN network switches instead of the proxy server, and I see reports of around 300mpbs regularly.

Are you using double parity on your Drobos? Initially, we had this configured, and saw horrendous speeds/reaction times when working with VMs... Drobo actually state that you should only use single parity when configuring a Drobo for use with a virtual environment.

Thanks

Dave

We aren't using the double parity on the appliances, just single. At this point I almost feel ripped off with the Drobo with that fact that the Synology devices seem like they performing better - even at a lower cost.

[murdocmk]

We recently implemented a 2 server vSphere 5 environment using 2 B800i's (one for primary storage, one for backup storage) and had enough problems with them to trade them out for QNAP TS-859s. My complaints were:

1) The performance/cost ratio was too low. An 8-bay QNAP costs significantly less than a 8-bay B800i, allowing us to buy more disks when we traded in for the QNAPs and still come in less than the B800i.

2) The "super easy" approach to managing the Drobo was not my style. I wanted more visibility into the workings/performance of the storage device. On the QNAPs, I can view OS logs, watch CPU/memory/IO performance and run SMART tests on disks.

3) I didn't care for the email alert mechanism on the Drobo. You have to have a system set up to monitor the Drobo using iSCSI (management LUN) and running the Drobo Dashboard software. The Drobo Dashboard software sends an email alert if there's a RAID or disk problem. It adds to the complexity (and thus likelihood of error) of monitoring the storage appliance. The QNAP can send an email alert directly from one of its interfaces. Sure, maybe I have to route that email traffic off the restricted storage network to get the email alert, but for small office setups that's not a big deal.

4) If you plug a USB cable into the Drobo it disables the Ethernet interfaces. I'm sure there's some technical reason for why that happens, but I think it's ridiculous. If an admin forgets that detail and connects to it using USB, they drop any iSCSI connections and pull the rug out from under their VMware environment. When used for backups that's not as big of a deal, but still silly.

5) The BeyondRAID thing might be cool for people that want to throw random disks in a storage appliance, but it just annoyed me. Even if I just have 4 disks, I may want to do RAID 10 in a small office virtualization environment to get the random read/write performance of RAID 10. Instead, I have to choose between either single or double parity BeyondRAID, both of which seem to have pretty poor random IO performance. If you think your sequential IO tests on Drobo weren't impressive, check out the random IO performance. In your case I realize that if you're using the Drobo more for backups you may not be as concerned about random IO performance -- although reverse incrementals, transforms, instant recovery and sure backup all benefit from decent random IO performance.

Anyway, I realize I probably went on a rant there about Drobo that didn't entirely apply to your situation. Just wanted to share my Drobo experience.

Mark

[btkrausen]

Thanks for that post...I'll never complain about too much information.

At this point, even though I feel ripped off, I'll likely keep the Drobo B800is since they are used only for backup purposes at remote sites. The running VMDKs live on storage blades within the chassis. Backups ran at a maximum rate of 78 MB/s last night in which the bottleneck shows: Source 99% > Proxy 26% > Network 0% > Target 11%, which tells me that my Drobo is running faster than the proxy can pull data from the datastores. However, when the backups run slower, say 39 MB/s, my bottleneck starts to point at the Drobo again: Source 55% > Proxy 21% > Network 16% > Target 83%.

[Dave-Departed]

99% of the time, my bottleneck points to source too. I can honestly say that we have no problem using the Drobo for backups, it performs fine. Hosting live VMs on it though, is probably a different story.

I'm going to set up a SureBackup job which used a Drobo volume as the datastore, so we'll see how that goes!