I already own a Drobo 2nd gen connected via Firewire 800 and today (05/31/2010) I have ordered a Drobo FS which I thoroughly will put through its paces. I will post all specs for my setup as well as the benchmark I am using so that everyone can compare his/her results.

While the BeyondRAID technology always was one of the main points to get a Drobo, the performance was far from stellar. The Drobo FS promises to have improvements for this issue. I get around 11-14 MB/s from my Drobo 2nd gen via FW 800 and I seriously hope that the numbers from Data Robotics for the Drobo FS hold true.

Update: If you can enable jumbo frames you get a better performance. Unfortunately Apple failed with the latest iMac Core i7 and Core i5 in not having jumbo frame support!

For those of you who just want the results, here is a chart with the results.

Drobo FS, Drobo 2nd gen Benchmark results

Read on for more details:

Test Setup

• 27″ iMac Core i7 2.8 GHz (iMac11,1), 8 GB RAM DDR3 (1067 Mhz), Seagate 1 TB HDD 7200 rpm (ST31000528ASQ)
• Drobo 2nd gen, connected via Firewire 800, Drobo Dashboard 1.6.8, Drobo Firmware 1.3.6
• Drobo-FS, Drobo Dashboard 1.7.2, Drobo FS Firmware 1.0.4
• Gigabit-LAN, Cat5e cabling, Cisco SLM2008 Gigabit switch
• for both Drobos: 3x Seagate Barracuda 1 TB HDD 7200 rpm 32 MB cache (ST31000333AS)

For benchmarking I am using the AJA System Test v 6.01 as recommended by Data robotics.

AJA System Test settings

I used the settings Data Robotics recommends, which are 1.0 GB file size, disabled file system cache, Video Frame Size 1920×1080 10-bit, File I/O API: Unix (pay attention to this!), round frame sizes to 4 KB and enable network volumes.

AJA Benchmark Settings

AJA Benchmark Preferences

First of all I benchmarked my network performance just to make sure my network is not a bottleneck. For this I started the benchmark on a second iMac (Core2Duo 3.06 Ghz, 4GB RAM, GBit LAN) and connected to my Core i7 iMac’s Seagate 1 TB HDD. Results over my local network were at 73.3 MB/s for write and 110.0 MB/s for read speed.

Local Network Performance

Drobo 2nd gen benchmark

Then I did some tests with my existing Drobo 2nd gen. I benchmarked it with about 60% disk usage and after I had reset the unit and its disks. Write and read speed where both at 14.3 MB/s which looks like there is some limitation in the Drobo hardware because usually the read speed should be higher than the write speed.

After the Drobo reset I got much better results. The write speed is now at 29.5 MB/s and read speed 34.3 MB/s. So it looks like the performance is significantly dropping the more data the Drobo holds. One thing I noticed is that in the Drobo Dashboard before it said „Drobo“ and after the reset it says„ Drobo disk pack“. I am not sure if this is anything performance related, just did not want to let unmentioned.

Drobo 2nd Gen 60 Percent Capacity

Drobo 2nd Gen 60 Percent Capacity Benchmark

Drobo 2nd Gen Empty Disk

Drobo 2nd Gen Empty Disk Benchmark

Drobo FS benchmark

The Drobo FS features 5 instead of 4 drive slots compared to the Drobo 2nd gen and it should contain beefier hardware since it sports a Gigabit LAN connection and supports Drobo apps. Data Robotics set expectations pretty low with a mentioned performance of 30-40 MB/s which already was in the range of the Drobo 2nd gen.

I had hoped for a little bit better benchmark values especially since the Drobo FS is a brand new Summer 2010 product from Data Robotics and the Drobo 2nd gen was released in Summer 2008. So there are two years room for their technology to improve.

Now on to the results. The empty Drobo FS had an AJA System Test result of 27.3 MB/s write and 39.7 MB/s read speed. While these results are nearly within the promised range from Data Robotics, the write speed is still somewhat disappointing given that the empty Drobo 2nd gen scored higher write values, a product which is over two years older.

I’ve also benchmarked the Drobo FS at 40% capacity, this should be enough to get some real world results to compare it to the empty Drobo FS. Here is the interesting bit, while the Drobo 2nd gen drops in performance if it contains more data it looks like Drobo FS is able to deliver a more constant performance. At 40% capacity the Drobo FS delivers a write speed of 26.7 MB/s and a read speed 39.1 MB/s.

Drobo FS Empty Disk

Drobo FS Empty Disk Benchmark

Drobo FS 40% Capacity

Drobo FS 40% Capacity Benchmark

Final thoughts

So was it all worth it? Well it depends, like always, if your main focus is speed then the Drobo FS is not the fastest NAS device you can get at this price range. The Qnap Turbo Station TS-509 Pro has much better performance at nearly the same price.

Why did I choose the Drobo FS? Well I have a business to run and while I am very techie I still do not want to spend more time than needed. The Drobo FS is a dead simple plug it in and get back to business solution. No bothersome configuration of IP addresses, formatting, choosing RAID-levels and expansion is as easy as pushing the next HDD in. It just works, out of the box with no hassles. The Drobo FS  is a second to none solution and excels at this.

Since I am an avid Mac user there is a second to none feature and that is Time Machine integration. Since firmware 1.04 the Drobo FS supports Time Machine volumes out of the box. Setup is dead simple, create a share, set a size limit for the Time Machine volume, let the Drobo Dashboard connect the volume and see Time Machine do its job.

Here is what I would like to see improved. While the performance of the Drobo FS is much more constant than the Drobo 2nd gen there is still a lot room for improvement. The write speed is not within the promised 30-40 MB/s, while it is no dealbreaker it is still not something I am really happy about. I do not know if  the hardware is not capable of delivering better performance or if the BeyondRAID technology is not yet optimized enough for delivering better performance.

While  I like the idea of the DroboApps there are not many really useful apps out there. Other NAS vendors have much more capable solutions. Apache without PHP and MySQL is not very useful, I could not even host a wiki on it. Since the performance is not up to par with the competition I am a bit doubtful if the Drobo has enough spare power to really host additional applications and still maintain its performance. For example a VPN-server, a Dropbox integration with selective folders would be something that would be highly welcomed. But both applications would surely need some CPU power. I am also looking ahead for the mysterious Oxygen cloud integration for the Drobo FS.

So for whom is the Drobo FS? It is for people who want an expandable storage solution that grows with their business It is for people who need a central shared storage for exchanging files and don’t want to buy a server just for doing so. It is for people who want a physically more secure Time Machine solution than a single external HDD. It is for people who want to focus on their business and not on how to maintain their IT.

Let me emphasize on this, Data Robotics BeyondRAID technology is something I have yet to see a match for, you always get the best solution with the HDDs you currently have. If you add more disks BeyondRAID adapts to it and you can switch between single and dual disk redundancy on the fly. So surely this is something everyone wants but would want let someone else worry about the details, which is what Data Robotics can deliver.

Update: Jumbo frames

Since I got asked in the comments if I could test the Drobo FS with jumbo frames, I did so. But, and this is really disappointing from Apple, all iMac Core i5 and i7 owners are left in the dust. Why you ask? Because Apple chose to use a network chip that does not support jumbo frames or at least did not enable it!

I have to do justice to Data Robotics if you enable jumbo frames on your NIC and switch you can achieve higher transfer rates. Write speed is at 34.8 MB/s and read speed at 44.3 MB/s if you can enable jumbo frames for all network components.

Drobo FS Jumbo Frames Enabled

Many applications are not designed to thread their tasks enough to really saturate the resources that modern multi-core systems offer. So software developers are encouraged to make use of technologies like Grand Central which takes a lot of work out of the multi-core software design process.

There are two arguments that justify buying multi-core especially quad-core systems today. Why you ask? Because benchmarking never shows reality, even real world application benchmarks only measure the performance of a single application on a multi-core system. But this is not how most peoples systems get used. We are not only running a word processor and do nothing else. We use a lot of applications at the same time, be it iTunes running with some encoding or playback in the background, be it your favorite browser with a dozen of tabs each with quite some Flash content and Javascript applications each taking CPU cycles in the background. Then there is Skype, a development LAMP-stack running in the background, Time Machine doing a backup and so on. So there are quite a few applications which hopefully do not take up all of your system resources but all of these running concurrently will make use of a multi-core system.

The second argument, we are at a point in time where multi-core systems have become a standard. So it makes much more sense for developers to take their time to re-engineer their software to make use of this extra power. The advantage is not always that things go faster but also that applications become more responsive while working on tasks in the background. This change will accelerate rapidly now that multi-core support in development tools and operating systems is broadly available.