Drobo

Report Your Drobo (V1, V2, S, FS, ...) Performance !

Thanks Jennifer, case submitted : #100715-000000

May be you could use some Ethernet Gigabit switch in front of your Mac Mini, which could give a better performance than the direct Apple Extreme Base Station connection.
Also, did you have other items connected to the AEBS during your 20 MB/s test ?
If some other connected item does not fully support a gigabit connection, or is faulty, that could explain the performance difference with your other test.

[quote]May be you could use some Ethernet Gigabit switch in front of your Mac Mini, which could give a better performance than the direct Apple Express Base Station connection.
Also, did you have other items connected to the AEBS during your 20 MB/s test ?
If some other connected item does not fully support a gigabit connection, or is faulty, that could explain the performance difference with your other test.[/quote]

No other device connected during testing. I´m copying some files right now and at the same time trying to watch 1080p content and that simply doesn´t work.
Im very interested in recommendations for another setup to gain speed.
How do you mean? Cable modem - AEBS/Wifi - (fast) GBE switch - Drobo-FS + Mac Mini?

AEBS = Airport Extreme Base Station wich is (supposed to be at least) 3 port GBE+Wifi+USB.

[quote=“DroboLars, post:22, topic:1597”]
Im very interested in recommendations for another setup to gain speed.
How do you mean? Cable modem - AEBS/Wifi - (fast) GBE switch - Drobo-FS + Mac Mini?[/quote]
Sorry, I meant Apple Extreme too.
What I was thinking is that may be your performance issue is solely due to the poor performance of the Ethernet hub within the AEBS, and if the only reason you connect your Drobo-FS to AEBS is because of the single Ethernet port on the Mac Mini, may be connecting both the Drobo-FS AND the AEBS to a fast GBE switch, itself connected to the Mac Mini, could improve your performance numbers.
It is just an idea worth testing if you can borrow such a switch, but if you cannot and it does not work, you could waste $50 without any benefit…

I think I can borrow one at work over the weekend an give it a try.
Sure would have been nice with 52mb/s.

Cable Modem <-> AEBS <-> GBE Switch <-> Mac Mini + Drobo-FS

[quote=“DroboLars, post:24, topic:1597”]Cable Modem <-> AEBS <-> GBE Switch <-> Mac Mini + Drobo-FS
[/quote]
More like :
{{Cable Modem <-> AEBS} + Drobo-FS}<-> GBE Switch <-> Mac Mini

I think we´re talking about the same:

Yes (impressive image ! :slight_smile:

An other idea : your performance differences may come from “Gigabit Ethernet Jumbo Frames” (9000 bytes instead of 1500 bytes), which apparently are supported by Drobo-FS and Mac Mini, but NOT by Apple Extreme Base Station (at least not by early models and/or by default, which is 1500 MTU).
You may search the Drobo Forum to learn more about that.
No idea if the latest AEBS hardware and/or any AEBS firmware upgrade can add this support.
And I don’t know either if having a 1500MTU-only unit mixed on the same switch with other 9000MTU units forces all dialogs down to 1500MTU (logically, it should, at least for low-end switches).

The test-results are without Jumbo Frames. JF slowed down the Read speed. I´ll try it once again this weekend when I get the GBE Switch up. (Edit: Wich supports JF).
If it improves speed I´ll buy one. Will this do the magic?

AEBS have a rep for being REALLY slow when it comes to sharing devices. Over on the Infrant forum, some folks are in a bit of an uproar because their shiny ReadyNAS NV+ is only getting 12-15MB/sec through an AEBS.

What are your speeds with the AEBS taken out of the loop?

It´s still in the loop to give me Wifi, but now connected as the image shows. I have now updated the testing/result post back on page 2.
Looks like I´ll be buying a JF supported GB Switch…

My understanding of what happens is that you get TONS of packet fragmentation and hence worse performance than if no devices tried to use Jumbo Frames.

Worst
Mixed devices Jumbo Frames and Non-Jumbo Frames
All devices Non-Jumbo Frames
All devices Jumbo Frames
Best

Well, the Dlink GB Switch does not seem to be significantly better than AEBS (Test #1 vs #4). And the Jumbo Frames look disappointing too with only +15% (Test #4 vs #6), although as bhiga suggests it could be due to the mixture of JF/non-JF.
Did you make any test with ONLY the Dlink GB switch between your Drobo-FS and the Mac Mini (AEBS unplugged) and with JF activated ?

[quote]Well, the Dlink GB Switch does not seem to be significantly better than AEBS (Test #1 vs #4). And the Jumbo Frames look disappointing too with only +15% (Test #4 vs #6), although as bhiga suggests it could be due to the mixture of JF/non-JF.
Did you make any test with ONLY the Dlink GB switch between your Drobo-FS and the Mac Mini (AEBS unplugged) and with JF activated ?[/quote]

No, it was connected as the image shows with both the Drobo-FS, MBP and the Mini Jumbo 9000 enabled, wich is the way it has to bee to get internet unless I replace the AEBS with a JF supported WIFI router (does that exist?)
I´ll find some time during the weekend to test single JF-units, but it looks like buying a new switch will give me aprox + 5mb/s.

Too bad the latest AEBS does not seem to support Jumbo Frames anymore than your 2007 model, at least according to its specifications on apple.com.
But may be you should check by explicitly asking, since it does not say either that it does NOT support them…
My understanding up to now was that “most” gigabit switches did support them.

According to smallnetbuilder.com it doesn´t.
This little Netgear-fellow does.

The tests are once again updated. I retested direct connection to MBP and this time I was not able to reproduce the 50+ mb/s, I therefore updated the results.
By summing up this I think we can say that this Drobo-FS is no where near the promised 30-40 mb/s. We can also agree that the AEBS is really not a speedy GBE switch. By using the D-Link Switch speed is up aprox 5mb/s. Jumbo Frames helps, but really not much.
Looks like this guys Drobo-FS is way faster than mine. This is my second unit. I guess I´ll contact support once again and hear what they have to say.

Disappointing that you could not reproduce your original 50+ MB/s results. Did you execute the original test a single time or multiple ?
It would be interesting to look at the graph of your results and see, if like mine, you have large performance variations from one frame to the next, and from one test to the next (different locations of the AJA allocated file within the Drobo block space ?).

This guy results on its Drobo-V2 filled up to 60% are very close to mine (around 15MB/s), and that seems to explain alone the 50% performance drop vs an empty Drobo-V2.
If true, my performance numbers would then be “normal”, except that they would be 50/60% below Data Robotics “best case” numbers, and show a disturbing “wearing down” property of the Beyond Raid algorithms… :frowning:

Average of 5 results. I have screenshots of the results, but not the graph. Tired of measuring now ;- )
I have filled it up a bit more so now its aprox 60% vs 50% when the first measurements were done.

Yes, according to that guy the Drobo-V2 behaves like that and that means your speed is normal. Have you not heard from support?

Here are the settings for Kona that we use for performance tests.

  1. Test: Sweep File Sizes
  2. Volume: Volume Name of Drobo
  3. Video Frame Size: DVCProHD 1080i60
  4. Make sure “Disable file system cache” is checked

Eeehh…then whats this: http://support.datarobotics.com/app/answers/detail/a_id/231/kw/aja/r_id/100004 ?

The link you refer to lists a special test for Drobo FS, which makes sense since it’s a network device.

FWIW, DVCPRO HD data rate is fixed 100 Mbit/sec as it originated as a tape-based format.