Feeds

Drobo B1200i: The heavy-duty array even your mum could use

It sure looks pretty, but can this storage box cut the mustard?

Remote control for virtualized desktops

Time to open her up: What's the performance like?

The review unit I have been working with is a B1200i with 12 drive bays. The unit shipped with 3x OCZ Talos 200GB SSDs and 6x Seagate 7200rpm Constellation ES drives. This left it with 3 empty bays for me to try out Drobo's BeyondRAID on-the-fly disk expansion. (If you don't know about BeyondRAID, Scott Lowe has a great writeup here.)

Performance is difficult to discuss with anything except vague generalities. The Drobo B1200i uses the three SSDs for tiering. Frequently accessed blocks are elevated to this transactional tier; it's incredibly difficult to reliably benchmark but does have a noticeable impact in real world use.

Without Tiering

Without tiering, the six 7200 RPM drives that came with the unit are cheerfully capable of sequential reading and writing at speeds that exceed that of a single gigabit interface. Divvying up the I/O so that it is more random does drop the available speeds below that of a single interface; two large sequential operations occurring each on its own network interface can achieve a sustained throughput higher than the max theoretical output of a single gigabit port.

Backing up files may will take some time (don't click to enlarge)

As with any spinning rust the more random the I/O, the more the performance degrades. This isn't unique to the Drobo; and the B1200i handles random I/O better than most. Hitting the Drobo with as many real-world workloads as I could generate (using two single processor servers each talking to a dedicated iSCSI NIC) I wasn't able to drop the responsiveness of the Drobo down below roughly 30 MB/sec for any given operation.

The box manages 100MB/s, including reading and writing, comfortably

Adding more spindles doesn't make the array any faster at the top end, but it does handle multiple random I/O requests better. With nine 7200 RPM spindles going, the unit provides three simultaneous sequential operations at around 60MB/sec each instead of the two that I get with six disks. The performance degrades more gracefully down to that 30MB/sec/operation stream real-world floor.

Unlike the six disk setup - where I was able to hit that performance floor using my real world setup - I actually had to throw some synthetic benches at the nine-disk setup to get it to degrade that far. I was also impressed by how well the BeyondRAID held up for performance during expansions and rebuilds; it was far faster than my Adaptec 3805s in a RAID 6 rebuild, ending up about on par with my LSI 1078s (PDF).

The straight-line speed of the BeyondRAID isn't as fast as my Intel RS2BL080 in RAID 6 on the same six disks. That said, the BeyondRAID handles mixed and random I/O better, and the B1200i doesn't have a 10-gigabit Ethernet port anyways, so the speed of any single operation is bounded by the NIC.

Overall, performance without the tiering is reasonable, better than most of the white box systems I could build to do the same job and comparable to any other eight or 12 disk entry SAN I've worked with around the same price range.

With tiering

Things get a lot more interesting when we start factoring in the tiering. Heavily transactional workloads such as databases see a huge boost. Frequently accessed blocks of databases saw up to a three times performance increase. VDI instances went from "okay" on spinning rust only to "actually quite zippy" with the tiering enabled. The tiering doesn't make a difference at all to large sequential operations; both because sequential data is rarely elevated to the flash tier, and because the spinning rust alone is capable of feeding the data at line speed.

The Drobo stretching its legs with some background tiering

In real world use I've run a domain controller, a moderate use exchange server with 100 users, a high transaction financials database (which was in the process of doing some fairly intensive integrity checks), 10 VDI instances, five Apache servers and eight image-rendering VMs all off the Drobo at the same time.

Initially, running this workload was significantly slower off the Drobo than off of the 3x two-disk RAID 1 array of Western Digital Raptor 300 direct attached storage setups that I had been using. After about a day all of the tiering had caught up and the overall package was not only smooth and more than usable, in the cases of the most heavily used items - such as the financials database - it was about three times faster than it had been before.

Setting up dual disk redundancy

With three SSDs, you might expect more of a boost from the tiering, but the Drobo is actually doing a dual-disk redundancy across the SSDs as well, so we're only getting the benefit of one SSD's worth of speed. For the loadout, the Drobo is reasonable. There aren't a whole lot of entry SAN units out there doing tiering, and the Drobo does a reasonable job of it.

The elephant in the room

I was initially quite concerned about OCZ as the SSD partner of choice. The failure rates on OCZ consumer SSDs have been shocking. I have personally had nearly every single one I've deployed (and most of the RMA returns) go bad - totally abnormal, even given the reported failure rates - and even as I write this I am concerned that this colours my objectivity.

I thought long and hard about bringing up this particular gripe; a review about Drobo is hardly an appropriate forum to air sour grapes about OCZ. The reality is that despite my personal aversion to OCZ and anything to do with flash storage, consumer product and enterprise product are not the same thing. Having talked to - and demoed - the B1200i for numerous other sysadmins, however, it is clear that the choice of OCZ is the elephant in the room.

Having worked with these disks for a month, I don't think the Talos drives are anywhere near as bad as the consumer level stuff that I suffered through. Drobo claims to have done intensive testing during the vendor selection phase. With several thousand of these units deployed in the field, they are not seeing abnormally high failure rates for the SSDs. It would seem that the prejudices against OCZ don't apply here. For the paranoid; the B1200i can do dual-disk redundancy across the SSD set. The chances of losing all three are pretty slim.

Real world testing

November and December are crunch times for a number of my clients; as a result they are also crunch times for me. This is the time of year where if something is going to break, it is going to break in a spectacular fashion where there are no spares and a replacement needs to be working now. In predictable fashion, the Drobo has been pressed into service as production item on my network and by extension the network of my largest client (with whom I share a data centre.)

The end result is that I have had to trust the Drobo with the business of my largest client during the busiest time of year. I'm pretty sure the read/write cycles on the SSDs are a little bit beyond what most reviewers are expected to put the equipment through.

Serving duty as the underlying storage for a rack's worth of very mixed-load virtualisation is not the market the Drobo was originally targeting, but it did a fantastic job of it. If you talk to Drobo's marketing folks, they envision the target market for this device as being creative professionals who need the ability to have large amounts of dynamic storage. Think a small office of video-editing nerds or photographers with Mac Pros and large storage requirements.

It makes sense; the Drobo is a high price tag device for the amount of storage that you get. Drobo sees itself as an alternative to products like the HP P2000, Dell PowerVault, and EMC VNXe.

If you have a technical bent and are willing to tinker a little you can go out and find (or build) an iSCSI SAN that performs more or less the same for cheaper. Synology has several options to choose from. If you want, you can easily go buy an LSI SA9280-16i4e for about $900 and build a server around it using LSI's CacheCade Pro that would absolutely wreck the Drobo for performance.

I however, see the Drobo as a serious consideration for several of my clients who have an obsessive need to be able to futz with their own networks despite knowing absolutely nothing about any of the stuff works. Here, I can provide them enterprise-class storage that even these folks can't break. These folks may not be on Drobo's radar today, but there's a whole market of tinker-happy pointy-haired bosses where Drobo is a great fit.

Conclusion

Drobo's pitch is that the B1200i the easiest to use block storage device ever made. If you need more storage add some disks. The display on the front will turn yellow until the automatic expansion is done, then turn green when you can fiddle with it again. If the unit is full and you still need more space, pull out one drive and put a bigger one in. Wait until the red light turns green before doing the next one. You can do this live. In short: you don't need to know anything about storage, iSCSI or computers in general to use this device.

This is BeyondRAID in action. Combine it with the dirt simple management software and I think Drobo have a large addressable market for their particular value proposition. People pay a premium for convenience all the time; Drobo is simply the convenient block storage provider. I can grok that. What's more, the convenience doesn't come at the price of a significant performance penalty.

This ease of use is why the Drobo has proven so valuable in the middle of crunch time for me. When the brown stuff hits the rotating air circulation mechanism, I don't have time to learn the quirks of a new device. The Drobo B1200i seems to have been carefully engineered to not have any.

On the downside, I am not friends with the management software and would much rather have a web interface. The Drobo doesn't have native deduplication, a shocking oversight for a premium offering, given that this is becoming a tick-box feature in the storage world. 3X 1-gigabit NICs are great for MPIO, but I would prefer it if there were a version available with 2x 10-gigabit SFP+ ports. Supermicro has managed to get the cost of 10-Gigabit switches below $300 per port; I expect 10GbE to start creeping into the SMB space in 2013.

In about a month of working with the Drobo B1200i, I have become completely and utterly dependant on it. Disentangling this unit from my network and building something reliable enough to replace it is going to be difficult. If I had the funds available, I'd be buying the unit off of Drobo, keeping it in production rather than sending it back. If you are in the market for an iSCSI SAN, Drobo has proven itself worthy of consideration. I give the Drobo B1200i a rating of 90 per cent; the few quibbles I have with it are minor. ®

Remote control for virtualized desktops

More from The Register

next story
Azure TITSUP caused by INFINITE LOOP
Fat fingered geo-block kept Aussies in the dark
729 teraflops, 71,000-core Super cost just US$5,500 to build
Cloud doubters, this isn't going to be your best day
Want to STUFF Facebook with blatant ADVERTISING? Fine! But you must PAY
Pony up or push off, Zuck tells social marketeers
Oi, Europe! Tell US feds to GTFO of our servers, say Microsoft and pals
By writing a really angry letter about how it's harming our cloud business, ta
You think the CLOUD's insecure? It's BETTER than UK.GOV's DATA CENTRES
We don't even know where some of them ARE – Maude
SAVE ME, NASA system builder, from my DEAD WORKSTATION
Anal-retentive hardware nerd in paws-on workstation crisis
Astro-boffins start opening universe simulation data
Got a supercomputer? Want to simulate a universe? Here you go
prev story

Whitepapers

Why and how to choose the right cloud vendor
The benefits of cloud-based storage in your processes. Eliminate onsite, disk-based backup and archiving in favor of cloud-based data protection.
A strategic approach to identity relationship management
ForgeRock commissioned Forrester to evaluate companies’ IAM practices and requirements when it comes to customer-facing scenarios versus employee-facing ones.
High Performance for All
While HPC is not new, it has traditionally been seen as a specialist area – is it now geared up to meet more mainstream requirements?
Reducing the cost and complexity of web vulnerability management
How using vulnerability assessments to identify exploitable weaknesses and take corrective action can reduce the risk of hackers finding your site and attacking it.
Saudi Petroleum chooses Tegile storage solution
A storage solution that addresses company growth and performance for business-critical applications of caseware archive and search along with other key operational systems.