Drobo

Is anyone using their Drobo S with USB3 ?

Just curious if anyone has switched to USB3 and what kind of throughput they are getting.

ahh is there a way to give an existing system (which currently has usb2) the usb3 feature?

Yes not w/ my DroboS but a Seagate GoFlex 3TB ext w/ a 3rd-party PCI-e USB 3.0 add-on HBA. It’s fasssst!

I’m trying to order a pair of DroboS 2nd Gen w/ the USB3 from my distributor and run some benchmark on Win7/W2K8R2 64-bit and MacOS X Snow this week.
Update from a channel partner there seems to be only 1 vendor selling MacOS compatible USB3.0 PCI-e add-on. Will try a 3rd-party on a MacPro tomorrow.

Looking forward to hearing your results rambo!

yeah, and if you do do reviews, would be fun to split them into a couple of parts.

Rambo, part 1
&
Rambo 1st blood … :smiley:

Using hacked in USB 3.0 support under OSX (Lacie has released drivers for their drives and the Mac community has made them ‘generic’ drivers). Just ran an AJA test:
Video Frame Size = 1920x1080 10-bit RGB, file size = 512 MB
Write = 80.9 MB/s
Read = 82.5 MB/s

This is using the built in USB3 ports on a GA-X58A-UD5 motherboard. My real mac does not have USB3 to test with and compare.

Whoa… For peace-of-mind, can you try like a 1 GB file? That would ensure we’re out of any caching the drives and/or Drobo is doing.

It seems the performance varies from run to run but:
Performance with a 1GB size:
First run:
Write = 76.7 MB/s
Read = 83.8 MB/s

Second run:
Write = 69.6 MB/s
Read = 74.7 MB/s

Third Run:
Write = 69.4 MB/s
Read = 76.9 MB/s

For some reason the first time I run the test it always gives high 70/low 80 MB/s writes, but repeated runs result in mid 60’s to high 70’s. If I shut down AJA and wait a while it is back to the higher numbers. Possibly some re-balancing/cleanup going on internally in the Drobo that slows it down.

After waiting a few minutes and running again:
Write = 79.7 MB/s
Read = 81.1 MB/s

Holy smokes, I am impressed!

does this mean that Drobo S simply has the bandwitdh as the bottleneck, and that by adding better cabling (eg usb protocl3) it allows more of the drobo S’s potential to be unleashed?

if so, could the same apply to an existing drobo v1 and v1 (which use USB 1.0/2.0 cables), if a 3.0 was supplied?

I highly suspect a faster processor and/or more memory is involved. The general bottleneck on Drobo transfer speed is the embedded processing of BeyondRAID.

ok thanks, i wonder if theres a pattern to the increase in processing, with amount of memory and cpu involved.
would be interesting to see a kind of linear/exponential graph from the DRI labs, such as those used on Moores Law

must be a limit where the extra benefit becomes too costly for what you get, or an actual physical limit where everything else cant process it as quick - but if its anything like that Moore Law, i think its been expanded anyway, whenever it seems to reach that limit :slight_smile:

Essentially it’d require an “internal” benchmark (without the external transfer port) to determine what the actual storage performance (BeyondRAID + drives) is like what ricardo’s done on the FS, compared to what the the external transfer performance is.

If I’m correct, the Drobo S with USB 3.0 should be able to saturate a USB 2.0 host connection compared to the standard Drobo S.

Of course some of it might be attributed to the change in USB interface hardware (the USB 3.0 interface hardware might be more efficient than the USB 2.0 interface hardware), but I would believe with USB 2.0 being mature, modern interface hardware should be quite efficient, so the contribution here would be negligible.

Which then leaves the hardware BeyondRAID processing itself to account for the difference.

Anybody have equivalently-populated standard Drobo S and Drobo S with USB 3.0 units? :slight_smile:

While that might be the case, I don’t think its responsible for the majority of the speed increase. I believe its the drastic bandwidth increase of USB3 that causes it.

The reason for this is that tests with the first gen Drobo S (like in this thread) show that almost this same speed is accomplished on the older units by using eSATA instead of USB2.

I believe this latest firmware update did more than just add >2TB support as well. Prior to the firmware upgrade I was averaging ~20-21MB/s over USB2. Now this number has gone up to ~26-27MB/s

eSATA is - of course - much faster, but unfortunately I can’t use it due to problems under Linux with AMD’s eSATA implementation in their 7xxSB chipset.

The Drobo S has a faster CPU than the Drobo v1/v2 though, so I probably should’ve qualified as “faster than Drobo v1/v2”