4

I'm looking at picking up a fully loaded drobo pro this year (last quater). Has anyone actually found any benchmarks on it yet? I've seen many people saying the speed of the current drobo mk2 is less than steller. I'd be probably looking to use it via the iSCSI interface though which I've anecdotally seen at +50% over fw800 speed.

Has anyone actual evidence of the speeds yet?

chewy_fruit_loop
  • 49
  • 1
  • 2
  • 6

8 Answers8

2

The short answer appears to be that there's no independent reviews yet, even the big Drobo lovers that are backed by Data Robotics are lucky if they have one Drobo Pro, so I think the answer is to wait.

There's also no good reason why iSCSI should be that much faster then firewire, as iSCSI has a lot of overhead, if anything 10-15% would make sense. Although firewire can be quite significantly affected by the chipset.

LapTop006
  • 6,466
  • 19
  • 26
2

Sorry for the cross-post, but I just posted about DroboPro iSCSI performance here:

Hyper-V and Drobo Pro

In addition to the GeekBrief.tv review with performance, there is also a really thorough review by the LA Final Cut Pro User Group with performance.

Regards,

Jim Sherhart - Data Robotics, Inc.

Community
  • 1
Jim Sherhart
  • 141
  • 2
1

I had the privilege of owning a DroboPro for just a few days. At first I was excited as though it was Christmas morning, and then, odd performance crashed the party like unwrapping a package of white knee-high socks. As an iSCSI target it performs decent when using the Microsoft iSCSI software initiator. However, I bought the DroboPro because it was VMware certified and using it as temporary / backup storage in my VMware environment was the intent. One, it didn’t work with VMware ESX 3.5 U3, which was the update I was running. And two, once I upgraded to a newer ESX, the performance was less than or equal to file transfer rates of my corporate DS3.

If I had more time at the office to work with the DataRobotic’s engineers, I think we could have gotten better performance out of the DroboPro. It didn’t work in my environment, nor many others out there, however, I would be surprised if DataRobotic's intentions were to push a product that didn’t perform as expected in their labs. If I had the money, I would love to give it another try.

1

I can tell you that at this time, at least with the two machines that I have tried, Xen and Drobo Pro don't play nicely. I was really bummed.

I have tried 3 different interfaces to the Drobo for iSCSI connection, all with the same results, Xen freaks out beacuse of errors with the data transfer. I am sorry for the lack of detail, I don't have time to get to my Xen test server and try it again. Got the same errors from 2 different motherboards, both AMD CPU, one Gigabyte board, other is ASUS (I think). Also tried from 3COM NIC, all showed the same error on the /var/log/messages file.

No problem using iSCSI from Windows XP with Gigabyte board, (same nic as one of the Xen servers I have) and from my Macbook Pro. Both yield sustained speeds of >20MB/sec write speed to the Drobo. Next on my list is to try it for Final Cut Pro and see what happens...

  • 1
    i'm looking to attach a fully loaded device for backup. The drobo seems to be a easy and cheep solution for this. At a push it may be asked to connect to vSphere, but it will more than likely by attached to a Win 2003 box via iSCSI on a dedicated gige switch, just to keep the traffic separated. I've never known windows to let you fully saturate a network link any way, and as I intend to use this for network backups, it may well be ok. running hd tune 2.55 on our PERC5 in raid 5 mode, gets an average of 84MB/sec is 25% speed of a local array good? I'm very green on iSCSI atm. thanks – chewy_fruit_loop Jun 17 '09 at 08:46
0

This article is the first result for "drobopro benchmark", but most of the links are dead. Using Xbench with 6x 1GB WD Caviar Black I got:

Sequential, uncached:
* Write 4k blocks: 35.5MB/sec
* Write 256k blocks: 31MB/sec
* Read 4k blocks: 3.5MB/sec
* Read 256k blocks: 50.2MB/sec

Random, uncached:
* Write 4k blocks: 3.2MB/sec
* Write 256k blocks: 9.1MB/sec
* Read 4k blocks: .58MB/sec
* Read 256k blocks: 10.75MB/sec

This is pretty paltry. By comparison, a single SATA drive got similar sequential performance (or better, 25MB/sec for the 4k read), and about double the random performance. I still really like the feature set of the unit though, so I instead just use it as a central disk for backups and archiving. For a primary file server drive though, I switched to a Highpoint Rocketraid and an external enclosure

Gabe Martin-Dempesy
  • 410
  • 5
  • 11
0

The regular consumer Drobo seems to cut the hard drive speeds in half according to this blogger's benchmark:

http://elliottback.com/wp/drobo-benchmark/

http://myownreality.eu/category/drobo/

Hopefully the Drobo Pro has improved performance.

rmwetmore
  • 432
  • 1
  • 5
  • 10
0

theres also some benchmarks and an unboxing of a drobo pro @ geekbrief #593

chewy_fruit_loop
  • 49
  • 1
  • 2
  • 6
-1

You need to be wary with the DroboPro. It slows down to a crawl as you add capacity. Ours ran well when we had about 8 TB, but once we went above that, we got tons of issues, including disconnects. I would not recommend it as a backup, unless it's really an archive. The reason is, the read performance becomes so poor that it would take too long to restore your data.

You may think this isn't an issue, but when you do need those backups you will be stuck- Suppose you're running an FTP server and you want to fail over to your Drobo- You will need to either wait 2 weeks for your data to be transferred to another medium, or go live with the Drobo and deal with 1 MBps of throughput for your FTP server. Neither's really an option. Also, the iSCSI disconnections would likely kill your uptime.

Simon
  • 1
  • 1