Update May 1 2007: This device started being flakey in fall of 2006 and became unusable in March 2007. Disappointing.
For background about our decision to try the Maxtor One Touch III : Turbo Edition 2x500GB external drive, please see the previous post about that. We'd been planning and thinking about this purchase/addition for over a year.
So we made the decision to buy the 600GB version of this first and didn't read the packaging really closely at Fry's. It turns out the 600GB version comes in two flavors, distinct between the packaging in just a couple of words, with one flavor only doing RAID 0, the other doing both RAID 0 and RAID 1. We purchased the wrong one, got it home, plugged it in, verified it was the wrong one and returned it. After some agonizing, we ended up buying the 1000GB version with RAID-1 because the main point is that we need data safety. Data is life.
The 1000GB (2x500GB) version was kind of expensive ($830 USD), but not ridiculously so per GB. It was just more than we needed at the moment by a couple hundred gig and thus a couple hundred dollars. We can use the space for extra backups. Further, we'd heard nasty things about the reliability of some maxtor 300GB drives, but we weren't sure if the drives in these units would be the same family (IDE vs SATA). However, the choice was made.
The first thing of note about these drives is that they are first multi-platform (not mac-specific) piece of hardware I've ever seen that prioritized OSX. The drive came preformatted for OSX and the manual has its OSX instructions first. Crazy.
The unit is very compact, just a little bigger than two small firewire enclosures jammed together. It runs pretty quiet, but there is a soft fan noise that is audible.
A couple of the reviews of this very new item complained that there was no way to verify that the RAID-1 was actually working. After we got it set up and running (pretty easy and quick), it became obvious that there was no way to verify that there was any RAID going on. On the back of the unit, there's a little tamper-evident sticker over one of the screws on the case saying that opening the case voids the warranty.
However, who the fuck cares about a warranty? I care about my data, period. I don't want the unit replaced if it goes bad: I want my data back. I'm not likely to send the unit with the drives in it off to get repaired if the drives go bad. If the brain in the RAID goes bad, at least I've still got the drives. If something like this went bad, we would open it up to try to fix it immediately, we're not going to put it in the mail and hope it comes back all worky with our data intact. You've got to be kidding me.
It is actually a little offensive that they put the sticker on there. I understand that this is a consumer level device and isn't built to be user-servicable, but this is also a RAID-1 device where the whole point of that type of machine is to protect my data.
We decided to try to open it up and yank the power on one of the drives and see if : 1) the device complained in an obvious way to let us know that something was wrong and 2) the device continued to work with only one of the two drives functional.
Every other RAID-1/5 we've had, we have done failure testing on to verify things work OK. Some of them, like the cheap motherboard built in SATA raids can take a while to rebuild and verify the drive that got out of sync, others are smarter (I've been very happy with the cheap PCI IDE Adaptect 1200A RAID cards and satisfied with the monsterous Adaptec 8-drive IDE raid card). But with this device, it was a big hassle to test the RAID.
So, despite it costing nearly $900 with tax, we peeled off the label and went to work.
It was not terribly easy to get apart initially. It was hard to tell what would break and what would give. It is a tightly assembled plastic case, with a number of little plastic prongs sticking down into the metal casing that made it hard to get apart.
Amusingly, double-sided sticky tape was used between the metal casing and the plastic sides to keep it all together well. That tape made it sound and seem kind of bad when we were pulling it apart, since it made it really tight and hard to tell when we were stressing the plastic parts.
After removing the two visible screws on the back plate, the largest sides (rubberized with MAXTOR printed on them) are squeezed forcefully and pried up a bit until the tape separates and we can see the plastic prongs sticking into the metal housing. Using a butter knife, we forced one of those to the side so that its hook would be free of the housing and then pried it until it came out. After that, the two large sides were pretty easy to remove, squeezing the edges in to loosen the prong/hooks and pulling firmly.
Next came the removal of the U-shaped dark-grey sides. This also has prongs that insert into the sides of the metal housing, but the do not pry out. The housing slides forward (towards the LED button on the front) about 1cm and the prongs are released. It is a little tricky and can require some bending to get the U-sides off, but the material is quite flexible.
After the plastic is removed, there are two screws on the metal housing that hold it together. These are removed and then the two halves of the metal casing slide and it comes apart.
Once the unit is apart, you can see the bright blue soft dampeners that reduce the hum and presumably protect the drives from shock a bit. The assembly is tight and seems well done. The SATA cables are easily accessible. We yanked the cable on one of the drives and slid the metal casing back together.
Plugging the power and firewire cable back in, the drive was recognized by windows and then popped up a warning message clearly saying that one of the drives had failed and we should request an RMA number from their 1800 number in order to return the unit for service. Woohoo. No beeping, but still, that's something. Hopefully that service isn't too CPU or memory intensive. I'd rather have beeping, to be honest, and no service constantly running in software. The unit should be self-contained as a data safe, I think.
Doing some basic testing of the drive, all the data appeared intact and everything worked. Basic RAID-1 test part 1 passed.
We unplugged it and then yanked the cable on the second drive, reattached the first, and tried that again. Same results, looked good.
We reattached both drives and tried it, no more nasty message, all the data looked fine, everything was working beautifully.
Reassembly was pretty straight forward: Put the case screws back in, put the U sides in place, slide the U sides on.
Stick the large sides back on, pop into place. The last little tricky part was getting the back plate on correctly. There are a couple of plastic hooks on the back and a place where the large sides attach to the back plate that need to get aligned properly. Not hard, once you see what's going on.
Notice the two plastic flanges at right angles to the main plate, in the center (from left to right) of the plate, near the edges, right under the fingers in that photo.
Once that gets slid into place, then you kind of need to pry the plate a bit to get the sides all aligned so they can snap together.
After it was done, it probably took us an hour from start to finish disassmbling, running the tests, photographing it, and putting it back together. Really quite easy once we got past removing the first large sides so we knew we weren't going to break the plastic.
Using a couple of windows drive benchmarks, using firewire 400 (normal), we got in the 35-36MB/sec range for reads. I have not yet tried it at firewire800 on my osx box.
Using ACDSee image browser (great product for windows), fire was able to snap through images as quickly as she could hit next and there was no noticeable difference between having the image archives on that box versus on her internal SATA raid drives.
Clearly Maxtor needs to do something about this to have a user interface (UI) that would allow a safety-conscious user to verify that the RAID is up and running, in sync, and everything is happy. It should also allow for an expert user to disable one of the drives for testing the RAID-1. Further, the box should actually make noise (simple SOS beeping, probably once every 15 minutes) if one of the drives is not working. This should be disable-able via a switch or software UI as well.
We copied the data onto the drive and started to use it in production as of today. We'll see what we think.
Posted by Earth at March 8, 2006 08:40 PM