Cooking with NDAS: trying out the Iocell NetDisk

What’s a little harder to find in either place is a more adequate description of what can and can’t be done with LPX, but I’ll come back to what I found about that topic shortly.

What was on XiMeta’s site was a later version of the LPX protocol driver installer, one that wasn’t flagged on the Iocell site. Also, back at the Iocell site I eventually found a more complete manual: the “User manual for NetDisk 3.3” is smaller than the online version of that multilingual pamphlet, but it walks you through the variety of ways you can link up your 351UNE, showing how the two NetDisk utilities are installed under Windows (XP only) and the ways you can stitch your NetDisks together into stripe sets or mirrors.

Not to plan

It’s also the first place that you’ll run into the occasional hint or direct warning that setting up this kind of more advanced network environment might not always go strictly to plan.

Only in this document does it become clear that NetDisks can’t be repartitioned or otherwise messed around with when more than one PC is connected to them.

What’s more, if you perform the smartest trick of “binding” multiple NetDisks together as mirrors or stripes – with the intention of creating a fault-tolerant or a very-high-performance collection of disks all treated as a single storage volume – you’ll run into another little limitation – namely, that it has to be a stripe or a mirror, full stop.

The penny began to drop for me that while there were some promising aspects to this product, it was very much a case of: not everything is possible all the time

The penny began to drop for me that while there were some promising aspects to this product, it was very much a case of “not everything is possible all the time”.

Serious downer

My first dose of impossibility was a serious downer: not only did the 351UNE not start up when I plugged it in, but it had no way of telling me what I’d done wrong, and the more extensive online manual wasn’t up to date, either.

Iocell appears to believe that the NetDisk will present itself via USB if plugged in that way, or on Ethernet if that link is live, but what I observed was that it did nothing on USB, and when I plugged in the Ethernet that wouldn’t go live until I unplugged the USB lead.

The instructions say that once the LPX driver and client software are installed, the NetDisk will present as another drive letter, but what I observed was that nothing happens until you go into the Disk Management part of MMC and find the volume that’s been added by the Iocell/NetDisk utilities.

In my test I’d taken the drives I was using from a pool of units, some of which had been installed in various PCs, others of which in NAS boxes.

The Iocell device doesn’t reformat the drives when it starts up – it just shows you whatever is on them and leaves Windows to partition and format them, and so several of my drives needed hitting with a big stick (that is, the diskutil command-line from which you issue a “clean” to wipe all their partition data) before they could be used.

Once I had a working drive visible via Ethernet through my Netgear GS724TR-managed switch, I thought I’d perform a rapid and worst-case test, especially considering that the test machine – my HP DC5800 – was running under Vista Enterprise.

I can hear the sharp intakes of breath all over the land, but hold your horses gentlemen: I’ve seen bursts of copy speed up to 80MB/sec from this network using Vista machines

I can hear the sharp intakes of breath all over the land, but hold your horses gentlemen: I’ve seen bursts of copy speed up to 80MB/sec from this network using Vista machines.

Bearing that in mind as a baseline, I had a go using the device setup just as it stood, without digging any deeper. I grabbed a heap of ISO images from my main server’s RAID and dragged them straight onto the single volume presented by the 351UNE, and Vista reported a steady 63-64MB/sec throughout the copy.

Traffic

Now have a quick think about where the traffic involved in this operation is going. My server is on one port of the GS724TR, the DC5800 PC is on another and the 351UNE on a third.

The server is throwing bits up the line packaged as TCP/IP datagrams, the workstation is receiving them as TCP/IP packets but then immediately sending them back out as LPX packets. (No flickering of the local machine’s drive lights was observed – this is solely network activity.)

Finally, the 351UNE is receiving them and stashing them on its own disk, and throughout this process performance remained comfortably above the speed range that I’d expect to see on a basic network. You can feel a “but” thundering toward us like the Severn Bore, can’t you…

But! Despite XiMeta’s and Iocell’s promises to the contrary (terse although those may be), that speed is nowhere near as “robust” as they’d have you believe.

Much as everyone might like to wish otherwise, the truth is that very few people with small networks who are looking to buy a £50 NAS box are going to see the sense in purchasing a top-end switch such as the GS724TR.

Netgear’s own blurb for this device says it’s a managed switch and that being managed is a useful thing for larger networks – as I’ve said here before, I’ve only ever tweaked one parameter on this switch via its management interface, and that was to turn on Flow Control.

Try again

So I thought I’d repeat my NetDisk test, but this time using a smaller, unmanaged switch – a tiny desktop Netgear GS105. Being impatient I just connected it via twin 1m patch leads, one to the DC5800 and the other to the Iocell box, then plugged the uplink lead back into a third port on the switch.