[Steve Modica]

When you test, open a terminal and run this:

sudo iosnoop -D

You’ll see each IO going to and from the device and the leftmost column will show you latency on the IO. I’d love to see what it looks like to an areca card these days.

Steve

Steve Modica
CTO, Small Tree Communications

[Steve Modica]

NFS was working great with 10.1, and then they released 10.1.1 and broke the locking.

With 10.1, you could enable locallocks on the clients and they would adhere to the lock files in the Library directories. The only possible concern would be if two clients tried to create lock files at the exact same moment.

With 10.1.1, apple started using NFS locking (using rpc.lockd) and that messed things up.
First, if you use locallocks, since the client can’t see the files are locked on the server anymore, it just blows through them and overwrites the locks. So clients will corrupt Libraries if they simultaneously access.
(You fix this by removing “locallocks” which makes sure they heed the locking)

Second, they seem to have broken things when a second client tries to access a locked Library. Now, the second client hangs. It has to be killed. Nothing gets broken in the library, but the hang on the client isn’t very nice.

I’m not sure if this is our implementation of rpc.lockd or what. I’d like to know if anyone has this working.
Our lockd is pretty standard.

Steve

Steve Modica
CTO, Small Tree Communications

[Steve Modica]

Advanced Format (4K sectors) is what you want. It’s better if you can use it. Not all cards support it, which is why the drives support 512 byte emulation. You don’t want to be using that since it’ll be worse than native 512bytes. So long as the card claims to support advanced format, you’re good.

Steve Modica
CTO, Small Tree Communications

[Steve Modica]

Have you tried an HFS+ formatted drive? I’m expecting it’s trying to generate extended attributes and getting an unsupported error so it’s barfing.

I would have a terminal up and capture the process with “dtruss -fp PID” and try to trace it right after you drag the file. I’d be interested to see what system call is failing.

Steve Modica
CTO, Small Tree Communications

[Steve Modica]

How “economical” it was 🙂

Steve Modica
CTO, Small Tree Communications

[Steve Modica]

The ThunderNET product is primarily so we can offer optical and larger port count devices. The SANLink is currently just 2 10GbaseT ports.

Coincidentally enough, both devices have all the same tweakable parameters! It’s very convenient.

Steve Modica
CTO, Small Tree Communications

[Steve Modica]

Since we sell the SANlink, I obviously like it better 🙂 I’m also a lot more familiar with the driver code.

Steve Modica
CTO, Small Tree Communications

[Steve Modica]

Bob! I was showing the darn things in the booth at NAB! You picked one up. 🙂

Steve Modica
CTO, Small Tree Communications

[Steve Modica]

There are lots of possibilities on a problem like this:

AFP can be set to “ignore ownership” in which case all the files will be owned by uid 99 (unknown to the server) and set to the UID of the person connected over AFP. (Samba users will see “99”)

Apple supports ACLs as does Linux. Check what ACLs are assigned to the files. (ls -le?)

The immutable bit may be set. Cameras often set this on media and it gets brought over when files are dragged in rather than imported with the camera’s software.

Steve Modica
CTO, Small Tree Communications

[Steve Modica]

I just have to add a reply and say I love Bob’s posts 🙂

As for thunderbolt networking, I had it setup for Kaku Ito in Japan so he could fly to Hawaii and take pictures of stars with 3 macs. It seemed to be working fine, but it was not quite as intuitive as I thought it would be (it bridged automatically and we had to edit the bridge and fiddle around to make it the way we wanted).

I’d like to say all of these vendors (including Small Tree) spend time working on performance, video editing and all the ancillary things that go with that (meta-data, sharing, latency, scaling, cross platform compatibility). For the most part, we aren’t just plugging crap together. There’s some promise of performance and functionality.

We spend time patching drivers and kernels, looking at how the apps like to read, and measuring latency to make sure things will work correctly (like FCP X and NFS, which seems to be a thing now).

I realize we seem more expensive, but we do an awful lot of work before we can get a box out the door and claim it works.

Steve

Steve Modica
CTO, Small Tree Communications