Is 10MB/s a normal write speed for a drobo 2nd gen on FW800?

Hey everyone,

The subject really says it all. I’m testing through the recommended Kona benchmark tool for OSX. My drobo is hooked up through FW800.

Some 256MB tests end up at 11MB/s, some at 9MB/s.

My Drobo is set up in two partitions. One 500GB Time Machine, and one that expands over the rest of the drobo containing media files. Writing to the larger partition seems faster, is that just me or not?


That seems slower than it should be. Make sure it is definitely running at FW800 speeds - http://support.datarobotics.com/app/answers/detail/a_id/98/kw/firewire/r_id/100004

It seems to be running at FW800 speeds. I have 4 drives in it, 2x1TB and 2x1.5TB. Two western digital green drives (salvaged from MyBook’s), and two other 7200RMP drives (of the same type) that shouldn’t be too slow.

Is there anything I can run over the drobo to optimize the filesystem? Defragmentation or simple repair permissions… or won’t that change anything?

Repair disk from disk utility. Also try a different cable and/or port.

Also do you have another devices daisychained off the drobo?
Do you have other devices plugged into the same FW bus?

Do not defrag the Drobo.
Disk Repair (Chkdsk, fsck) OK, but don’t defrag!

Thanks for your answers guys. I won’t defrag it, I’ll run Repair disk from disk utility again. Nothing else is using the FW800 bus, and nothing is daisychained.

I have another problem with the drobo, that might be linked to this one.
As said in my first post, I have one 500GB drive for Time Machine. When time machine does it’s initial backup to the sparsebundle on that partition, everything is okay. But then at some point it corrupts the sparsebundle, leading to an error 109 when it tries to mount. This usually happens when it’s been doing incrementals for a few days.

I’ve read some tutorials on how to fix the error, and I’m able to follow those steps and regain access to the sparsebundle. Only to have time machine corrupt it again the day after :frowning:

Time Machine shows this behavior when backing up to the Drobo when it’s hooked up to my Airport Extreme base station. So I’m not sure wether the problem is the Airport Extreme, or the Drobo raid array, or the combination of both.

This happens when I let Time Machine create the sparsebundle, and when I let it back up to a sparsebundle created through the automator script floating around in this forum. BackUpMyFruit I believe it’s called.

Perhaps there’s something wrong with the initial format of the Drobo disk pack, that causes the backup to go corrupt and transfer/write speeds to be slow?

Is the Mac (being backed up) connection to the Airport Extreme wired, or wireless?

When I’m working at my desk it’s connected through wire (gigabit speeds), sometimes it backs up through wireless.
I’ve read on another site that the problem is the airport extreme, and that a fix would be to connect the drobo to another mac. Shame my mini is the media center, I don’t want to have the drobo making noise in the living room. And I’m doubtful that’ll solve the problem.

If both wired and wireless connections are simultaneously active (and not some kind of automatic switching), you may want to see if you get better results with just one connection active, though I do suspect the Airport Extreme as you say.

You shouldn’t be having data corruption - a dropped connection should just result in an incomplete file operation. Unless the Airport Extreme is rebooting under high load?

I believe the Airport Extreme just gives up when it’s processing multiple requests to the network share. I keep the drobo next to my mini in the living room now. I’m finding the noise acceptable, Time Machine isn’t corrupt (yet…). Write speeds haven’t improved.

Now I’m doubting (yes, I came up with another question to bother you guys with :wink: ) wether I should or shouldn’t run drobo dashboard on the mini. E-mail notifications can be nice, but since it’s divided in 2 partitions, the drobo dashboard can only see the first partition. It therefore sees the drobo as less than 10% filled, and adjusts the lights on the drobo accordingly. (My drobo is 40% filled).

Aside from missing e-mail notifications. Is there any other downside not running the drobo dashboard? I check for updates regularly, manually.

None that I can think of - Drobo Dashboard isn’t required to use the Drobo.

Alright. I really can’t take it anymore. I’m at a friend’s place, he’s currently writing to his drobo at 24mB/s.
I’m going to reformat my Drobo. Is there anything I should think of, in order to maximize performance?

I’m going to format to HFS+. I would like to have one 500gb partition for time machine, but I can do away with it if it would be better to have one large partition, performance-wise. Also, is it better, performance-wise, to choose 8TB for the expanding partition, or can I just choose 16TB?

Do more tests first… I say this because even directly connected (via USB 2.0), my Drobo gets varying performance between 5 MB/sec and 30 MB/sec. A lot seems to depend on the file sizes involved.

On thing that comes to mind - is the Time Machine partition always/frequently in use? Since it’s two volumes on the same physical set of disks, I’m wondering if you’re showing lower performance because the OS is accessing both partitions.

I run the file transfer tests whenever the Time Machine partition isn’t used. Of course, when that partition is active, the write speeds would drop even more. File transfer speed always starts out at 20 or something, and drops down to 10 within the first 15 seconds.

So to summarize: If I decide to format the whole thing. Can choosing 16TB volume size (in stead of 8 or 4) negatively affect peformance? I’m not worried about boot times, just read/write.

I’ve reformatted the Drobo and put all my data back on it. I’m now getting write speeds around 30mB/s and read speeds around 47! :slight_smile:

Thanks for your help bhiga and others.

Interesting… I’d keep checking performance regularly to see if the slowdown was just some weird fluke, or if it relates somehow to the content (number of files, or space used) of the drive.

I’ll keep checking over the coming weeks. The data that is on it now is almost the same as before, about 100GB less, those 100GB were mostly large files (dvd rips).

Perhaps the slow performance was due to the following: When I first bought the drobo, I put two drives in it, thus automatically creating a raid 1 (mirror) setup. Correct me if I’m wrong here, but I believe this is how the Drobo works. Then I upgraded to 3 and 4 drives, turning it into a raid5 array (BeyondRAID I believe the Drobo’s algorithm is called). Perhaps something went wrong in this process, causing the Drobo to slow down.

On the other hand, It might’ve been the 5 or 6 time machine sparsebundles that went corrupt. Each of them contained about 900.000 files after the initial backup…

Whoa… yeah, I think your thinking is correct - either the Drobo never finished the relayout to the expanded volume or the filesystem started suffering due to the corrupt sparsebundles.

I had the same experience. I have two 1.5TB drives in a 2nd gen Drobo running on FW800. I was seeing terrible write speeds (less than 10MB sometimes and other times it would just stall completely). I re-formatted and put the data back on and performance came back to normal. Now after a few months (same drive configuration), my performance is terrible again. So . . . keep an eye out and post back here if you have performance problems again. I’m frankly getting a bit frustrated.

When was the last time you ran repair disk or chkdsk?