The Drobo was trivial to set up. The front cover pops off, and you slide the hard drives into the bays. The drives were a little tricky to slide in—they stuck a little until you got them in far enough and everything would start lining up properly. Once in there’s a catch that clips around the front of the drive to hold it securely in place, and that catch also serves as the eject button when you need to remove a drive.
The mSATA SSD card was also quite easy to install. Flipping the Drobo 5D over, there’s a panel on the bottom of the unit which removes without tools, exposing the connector into which you insert the mSATA card. The 5D recognizes the SSD card automatically and uses it for caching.
I installed my new USB 3 card in my editing computer, loaded up the driver for it, installed the Drobo Dashboard software, connected the USB3 cable, and turned on the Drobo.
The boot sequence of the Drobo 5D takes a couple of minutes. The Drobo was recognized by my editing computer, but when I tried to use the Drobo dashboard to format the unit, the format failed. I then tried to update the firmware of the 5D to the latest version, and that also failed. So, I decided to try this with my Lenovo W530 laptop.
During the Drobo boot sequence when connected to the laptop, I received an error message about a new device being found, but the device driver failed to load, and the Drobo was not recognized by the computer. Not good. I rebooted my laptop and tried again – still no luck.
I called up Drobo technical support, and was quickly connected to a technician. Over the next 45 minutes, we tried a variety of things, and found that the 5D worked fine when connected to a USB 2 port on either computer, but not with the USB 3 ports. Eventually it was realized that the USB 3 cards from Sonnet are not supported by Drobo—Drobo technical support said that there are implementations of USB 3 out there that do not fully support the USB 3 standard, and Drobo requires a completely implemented USB 3 interface. Drobo technical support sent me a link to a posting on their support site that lists supported and unsupported USB3 interfaces. In general, however, Drobo suggests a USB 3 implementation based on an Intel chipset. So, the Sonnet card was returned, and I purchased a card on the supported list manufactured by Buffalo.
Using the USB 2 port, I was able to update the 5D’s firmware to the latest version, and format the drive into one 16 TB volume.
With the Buffalo card installed, I booted up the Drobo 5D. Shortly into the boot process, I heard the Windows 7 USB sound indicating a device was found, but then got an error about the driver failing to load, and the Drobo 5D was not recognized. Sigh.
With a little trial and error, I found that if I waited until the Drobo completely booted before connecting the USB3 cable, my editing computer recognized it with no problems. And, if the Drobo is in standby mode when the computer boots, the Drobo boots at the same time, and the computer recognizes it without a problem once it’s booted.
In further consultation with Drobo technical support, we realized that my motherboard did in fact has USB3 functionality that could be turned on. Using the built-in USB3 port, the 5D was recognized without a problem then it booted. On shutdown, however, you get a message from Windows 7 popping up that says “volume needs to be formatted before use”. I did find, however, that the performance of my internal USB3 port was poor compared to the performance of the Buffalo USB 3 expansion card, and so have chosen to continue to use the expansion card.
And just recently, Lenovo released updated drivers for the W530 laptop including those for the USB3 port, and with these new drivers, my laptop recognizes the Drobo. I get several connect/disconnect USB sounds from the laptop as the Drobo boots as well as an error message about a USB device not being recognized and having malfunctioned, but once the Drobo 5D is fully booted, it works with my laptop.
While Thunderbolt is a more commonly found port on Macs, the Thunderbolt interface isn't widespread amoung Windows systems. Being a Windows guy, I did not test the Thunderbolt interface.
The latest firmware release also includes the ability to create a dedicated “backup” volume of a fixed size, so that you can use the Drobo 5D as backup storage without having to monitor it to prevent it from consuming your entire primary volume.
Once all these hurdles were overcome, it was time to start experimenting with the system. I did some benchmarking both with the SSD card removed, and then with the SSD card installed, to see if a difference was noted.
Before I present the results, however, it’s important to point out that because of the way the SSD card is used, benchmarking is problematic and any results should be taken with many grains of salt. With the Drobo 5D, you should see an increase in read speed on the second time you access a file, not the first, as the first time you access a file on the 5D is when the Drobo also loads the file into the SSD cache. If the benchmark only reads its test file once, no difference is going to be seen. So, if a benchmark is not showing an increase in performance, it may be completely misleading depending on how the benchmark program operates.
Without the SSD cache card installed, the ATTO disk benchmark gave write times ranging from 81 MB/s to 87 MB/s for block sizes in the range of 64 to 8192 bytes. Read times for the same range fell between 36 MB/s and 111 MB/s.
With the SSD card installed, write times were consistently 86 MB/s for the same range, and ranged from 85 MB/s to 105 MB/s.
The increase in speed with the SSD installed for the ATTO benchmark ranged from no difference at all to 175% increase depending on the block size. Three block sizes (32, 64 and 128 KB) actually showed a decrease in performance ranging from -8.6% to -14%.
I also ran the Roadkil.net Disk Speed benchmark. This benchmark measures both linear and random disk reads of various block sizes. For linear read, with block sizes ranging from .5 KB to 1024 KB, the difference in speed between without and with the SSD installed ranged from -72% decrease in performance to +202% increase. 9 out of 12 block sizes showed a decrease in performance with the SSD installed.
On the random access side of the Roadkil test, the performance change from without the SSD to with the SSD ranged from -55% decrease to a significant 1,016% increase. The block sizes from 0.5KB through 64 KB experienced a minimum of 536% increase.
Given the nature of the SSD cache, I also decided to write my own benchmark that would perform multiple read passes through the data, as well as write and read a separate new data file of same size as the amount of memory installed in the computer to clear out the Windows 7 file cache in between each read pass of the test file. The results I obtained showed a 2.7% to 6% increase in read performance in the second through fifth read passes.