McBeen.com

Ten’s Complement finally got back to me today, and added me to their beta tester list. I’m not sure what that means for now, but hopefully I can get access to their new products and put them through their paces. supposedly they are doing a full ZFS platform, including my favorite technology, reduplication.

I have been considering upgrading my old infrant chassis for a while now. It works great, but it’s painfully slow. Its been efficient and very reliable for me. Netgear continues to update the firmware to this day with new features and bug fixes. It has been a great investment.

That being said, spending 600 dollars on a new chassis is more expensive than actually building a full machine and tossing the drives in it. Believe me, I fully understand the cost tradeoffs here. I understand there is a lot to be said the the dividends and time savings on the backend. Still, I have held off.

Then, this christmas the family was especially generous. I picked up a few Newegg gift cards, and decided to redeem them for a Thermaltake level 10 gt. It has a ton more space than my original chassis from over two years ago, along with numerous hot swap drive bays.
I had a few drives kicking around, so I tossed them in the extra slots and played around with setting up a software array in OS X. It worked out fine, but the redundancy options were limited to RAID 1. I looked around for something more interesting.

Today I came back across an old project I had completely forgot about, ZFS. If you are not familiar with it, you should go look it up. Its possibly one of the best fundamental technologies out there for hard drive systems. Apple was working on embedding it into OS X for a few years, but had to abandoned the idea after licensing fell through after the Oracle/Sun Merger. I found there was an open source port of what apple had been working on, and that a prepackaged installer was available.
I went ahead and downloaded it and tossed my spare drives at it. Soon after, I had a nice large pool of raid5 equivalent storage just beckoning for my data. I started tossing files at it and was quite surprised at the performance. Using 4 drives I’m pushing over 200 mb a sec through the array on write. Considering this is all running in software, its pretty good.

ZFS supports tons of advanced features, and easily allows for in-place expansion using larger drives.

I may have finally found a very interesting infrant replacement.

» Re-using an LCD screen JeeLabs.

The box arrived today. I’m honestly surprised. I was half expecting the place to be a front for stealing my credit card information.

But in reality the size was per spec, and in great shape.

I took about a day and reverified all my measurements and scales. I went over my cad drawings again and made sure that all my planned cuts were good. The next morning I got out my hack saw, drill and files, and began to grind away on the block.

It took a few hours, but I was rewarded with a perfect fit. The reason for the over sized chunk of copper incase you are just join me here, was the original water supply cover for the block was not the correct shape for my power module. I purchased a chunk of copper to match the original mounting surface as close as possible.

I then set the water supply cover over the block, and went pale. As it turns out, for reasons unknown, there turned out the be a space between the water block and the circuitry. It was about 3 mm too short, even though the copper block was the exact same height as the original. There must be a design spec I missed somewhere. I have two alternative at this point, and neither are all that great. The first is to order a new piece of copper thats slightly taller than I need, then grind it down. That means more money, and more time grinding on an even thicker chunk of copper. The second is to fashion a spacer of some kind to go between the water supply and water block. I have plenty of plexy around, and its easy to shape, so I went with that first.

I took a 9mm chunk and clamped it to the copper block, then used my radial file to rasp it down. It went much faster than I expected.

After that I cut a water channel and epoxied them together.

After setting the block in the 4870, I measured the required height for the spacer, to make sure 3mm would function properly.

For the next part, I had to remove about 5mm from the top of the spacer, while keeping the surface perfectly flat. I worked on it in sections, using the radial file in my drill. I ground the spacer down to a rough 4mm thickness, and then reverified spacing on the card. After making sure 3mm would work, I went to work with my flat file, and honed it perfectly flat.

I then cleaned all the parts with rubbing alcohol, and epoxied the finished block to the water supply, being very conscious of the tight spacing required for assembly.

After everything set-up overnight, i used silicon adhesive to glue the cover plate back onto the top of the water supply, and clamped it down to dry. I left that for 24 hours to set-up as well.
The next morning I added some water to the block and held it vertically to see if I had any leaks. As I watched, water began to slowly drip from one of my corners. I had to cut much of one of the corners out to do a capacitor placement. It appeared that the cut was undermining the effectiveness of the joint in that area. I decided to see if I could strengthen it, and over layered the section with more epoxy, allowing it to flow into the crack. I then let that set, and tested later that evening, only to see the water still flowing. It was at this moment that I realized the water block was not really attached to the water supply plate. I was able to pull it off with little to no force. The epoxy was unable to form a bond with the water supply, although it bonded to my plexy just fine. A that point, I decided to go ahead and switch to silicon rubber for the entire build. I cleaned up all my parts, and started over.

As of tonight, The water block has passed its first water test. It was able to hold water under no pressure for an hour.

I’m still very concerned about not having any mechanical support for the silicon joints. I’m still working on tapping the block for 6/32 thread screws, but so far I have not been able to find the right drill bits for the process. I have self tapping screws, but I need a precise drill bit size to make it work. As soon as I find a set, I’ll be adding screws to make sure no leaks form after months of use and aging.