The family mainframe (by )

I'm in the process of consolidation the home fileserver and the public Internet server - currently two separate bits of hardware - into a single physical device, virtualised to support multiple indepedent machine images. Having a single family mainframe will simplify the management of the complex web of computers and services that support our digital life.

For various reasons, the best place to build such a thing is at the office end of my workshop. Even though it's at the "clean" end, this is still a room that is prone to having fine conductive dust in the air, varying humidity and temperature, and (heaven forbid) a leaking roof. Also, I want a case with extensive room for upgrades, and which makes it easy to replace parts. Having used 1U rack-mounting servers for quite some time, I am sick of highly compact servers that are difficult to work with, requiring extensive dismantling to get to parts.

Clearly, I needed a rather special chassis for this new family mainframe, so I bought a load of steel, picked up my tools, and got to work. I've been working on this for months; I initially cut up the metal at home, then visited a friend's workshop to borrow his pneumatic rivet gun and his MIG welder. Since obtaining my own TIG welder, I've been able to continue at home.

The chassis is nearly structurally complete; this weekend, I've been attaching mounting brackets inside it for everything to attach. All that remains is to finish welding the upper panel on, then the whole thing can be cleaned and galvanised, and the exterior painted. Then I can fix it to the wall and start fitting the electronics and electrical systems!

The first thing I did this weekend was to fit mounting brackets for the processor frame. This is taken from a standard ATX case, and is the base plate with standoffs to mount the motherboard, the frame to attach expansion cards to, and the frame to hold the PSU. This is screwed into the chassis, so that I can use an existing frame (rather than having to make one myself), and so I can replace it if needed. The frame is held in place by two locating pins that fit into holes in it, and then two screws through the upper-left corner (I drilled and tapped holes in the top left bracket), and a little spacer at the top right to stop it from flexing:

Processor frame mounting brackets

With the frame in place, it looks like this:

Processor frame in place

Next came the expansion frames. I may need to add additional hardware inside the chassis in future, but once it's holding a running server and painted and bolted to the wall, I can't really take it down to weld additional brackets into. So I cut off one-inch lengths of square tube, drilled and tapped a hole in the centre of one side, and welded them to the inside of the chassis. I drilled holes in the ends of strips of steel, so they screw into the pairs of brackets, creating a metal strap that can be removed, things mounted onto (via welding etc), and then screwed back into place, without causing major disruption. There are two - one beneath the process frame, above where the UPS will go; and another right at the top, above the environment management system.

Here's the upper one:

Upper expansion frame

And here's the lower one:

Lower expansion frame

The welds were quite difficult, as I had to reach right down into a corner of the chassis. As such, they were either OK or awful, depending on whether I had to use my right (dominant) or left hand:

Lower expansion frame (left hand bracket)Lower expansion frame (right hand bracket)

I also cut and drilled some mounting flanges, which will be what are used to bolt it to the wall:

Mounting flanges

When I made the sides of the chassis, I welded angle iron onto them, in order to attach said flanges:

Tabs where the mounting flanges will attach

(Note the plasma-cut hole, which will be where a removable plate with sockets for Ethernet, VGA, and USB will go).

The mounting flanges are quite thick (the wall is rough and bumpy, so the chassis needs to be spaced slightly from it), so it was good fun welding them to the much thinner angle iron. I think I did an OK job:

Mounting flange attached

Then I mounted the internal frame for mass storage devices, which goes above the processor frame, below the environment management system. It's a metal plate drilled for lots and lots of 3.5" disk drives, which attaches (with screws) to brackets I welded into place:

Mass storage frame

With all the internal stuff done, I started to weld the top panel in place, which I'd avoided in order to enable me to get access into the top:

Top panel

Annoyingly, I ran out of argon while doing the tack welds. A TIG welder without shielding gas is a lot like a plasma cutter, and I burnt a nice hole in a shower of sparks. It's only a small hole, so I'll be able to weld over it when I finish the job off.

Unable to do the final welding, I drilled a hole in the eaves, where clean outside air will be drawn in through a duct into the environment management system:

Air inlet

I also hefted the entire thing up to the wall where it will be mounted, propped it in position, leveled it, and drilled through the holes in the flanges to make the holes that will be used to anchor-bolt it in position:

Wall prepare for mounting


  • By @ndy, Mon 7th Sep 2015 @ 9:56 am

    Great work! 🙂

    How will you stop moisture from coming in through the air hole?

  • By Alaric Snell-Pym, Mon 7th Sep 2015 @ 10:34 am

    The amount of water air can hold without it condensing (into fog, or onto surfaces) depends on how hot it is, so the incoming air will be heated by the "environment management system" alluded to. I'm not yet sure how much heating it will need in practice - which is why the EMS also includes temperature and humidity measurement and logging devices!

    One danger is, after a cold night causing the metalwork the air immediately impinges upon being cold, a change to warmer moister air, which then goes onto cold metal and is cooled down and condenses the moisture it then can't hold any more.

    The other danger is outside humidity being so high it actually becomes fog, which will then stick to surfaces and make them wet.

    Either way, the EMS design includes a space for an old CPU heatsink in the air stream, onto which I plan to fix some high-power resistors to make sure the air has been warmed up on the way in!

Other Links to this Post

RSS feed for comments on this post.

Leave a comment

WordPress Themes

Creative Commons Attribution-NonCommercial-ShareAlike 2.0 UK: England & Wales
Creative Commons Attribution-NonCommercial-ShareAlike 2.0 UK: England & Wales