New isn’t always better in creating an Enterprise class network infrastructure in your own home. I call this concept Enterprise Home Computing, a term I define as using previous generation Enterprise class hardware in the home or home office environment.

One of the chief costs involved in setting up a home network where you are also hosting your own e-mail, web, or other services is computing hardware. Complicating this is the personal cost of time and out of pocket expense when you have a sudden system failure since you may not have the luxury to have a dead server sit for two weeks while you wait for your next paycheck (that e-mail you missed may have been a contract job lost).

So what do you do? Well, you buy name-brand hand-me-downs. Seriously. Keep reading to learn more…

Companies like IBM and Compaq didn’t fill data centers with racks full of high priced servers because their equipment suffered failures every other week. Also, someone paying prices starting at $10,000 for a single server … and they might be buying hundreds of them … want choices in OS selection, because some companies believe it or not, don’t run Windows on their servers.

Enterprise class systems are generally light years ahead of consumer grade hardware in the areas of stability, reliability and hardware configuration. Even an older Enterprise class system can be competitive with a modern consumer grade system depending on the application.

Additionally, when you’ve paid that much for a server (much like buying an expensive car like a BMW), you also get a certain level of service and support for a fixed period of time. For instance in the case of IBM, at one point certain server class systems were warranted for 3-5 years from the date of sale, irregardless of their current owner.

The corollary to that is that when that SLA (Service/Support Level Agreement) or warrantee expires, many companies consider that hardware to be a risk as they may face uncertain costs to repair failed hardware that is unsupported by the manufacturer.

What do they do? In many cases over the last decade or more, they call their friendly local computer recycling companies and ask them haul it away. These recyclers sometimes turn these systems back around on eBay (either whole or parted out), refurbish them to donate to charities, or break them down to reclaim precious metals to name but a few possibilities. Many of these companies that sell on eBay also have pseudo store front operations where the tech savvy can browse their stacks of incoming receivables and haggle over untested/unknown hardware (here’s where doing your homework is invaluable). Not all sellers let you do this though so it may not be an option in your town.

The other half of the time, big companies run these servers till they burn out (figuratively speaking) and this can also be your friend as just because a PSU (Power Supply Unit) has failed on a server, depending on the nature of the failure other parts may still be viable such as memory, harddrives, adapter cards, etc… Even something as simple as a clean chassis (with operator keys!) can be value to you.

Also, when several businesses each buy 10,000+ computers and run them till they drop. Odds are that they are going to have the leverage to demand ongoing driver support for future OS versions. This means that a six year old hardware architecture designed for an OS from that time period might very well end up with current driver support two OS generations later. In the consumer hardware space this is almost impossible… and if there are that many systems in the field, you can almost guarantee someone is maintaining current Linux drivers as well. Remember Google is your friend here. Search the name of the server model along with the term Linux and see what distributions come up in the search results as having success.

The reality is that a well maintained server (run on protected power, kept dust free, etc…) can literally run for years beyond EOL (End of Life). In almost all cases, your biggest enemies are power damage, over heating and harddrive failures. Frequently the first two cause the later.

So what do you do? Study up on what was hot in the market 4-6 years ago and start your eBay searches. Your sweet spot is $100-$150 for a fully loaded decommissioned server and when you find it you’ll literally see dozens, if not hundreds of them listed for sale. Want to build a cheap second hand data center? Sometimes you can get pallets of systems at almost pennies on the pound.

Next step? Buy two or three matched systems. This achieves two goals. First off, it gives you spare parts on hand since five years from now you’re not going to be able to find this hardware anymore… especially at these prices. Second, it gives you materials that you can consolidate from. Here’s a case example:

You have a server that has four harddrive hot-swap bays and typically came stocked with two HDs. The system also has four to eight RAM slots on it’s motherboard and again may have typically only come stocked with half of those slots holding memory. If you buy two systems, you may be able to consolidate all of the RAM and HDs into one server and suddenly you’ve taken a system that might have sold used for $500+ in this newly arranged build configuration for less than $300 (not counting shipping). You’ve also covered you major failure point in that you have a spare motherboard, RAID controller and PSU in the event that one of them fails. The only thing left to do is purchase a few more stray harddrives that match the specifications of what you now have and you’re ready to rock! Keep a sharp eye, sometimes HD resellers will actually have cases of legacy drives for sale. It might be cheaper to buy 10-20 refurbished HDs in a case than 4 new drives… and at that point you can have a couple of failures and you’re still ahead on cost.

Some critical reminders as well. Before you make that purchase, double check that there is really driver support from the manufacture or OS developers for your target system to support the OS you want to install. Then download everything immediately to an archive folder incase it disappears tomorrow.

Next when you get the system put together, make sure to install all of the latest BIOS and System Level patches for all of your hardware (you downloaded these to your archive folder right?). This will insure best compatibility and stability with current generation Operating Systems.

Next, if you happen to have it, run full system diagnostics to test memory, CPU, etc… 99 times out of 100 this is not necessary, but when you do find a problem you’ll be glad later that you dealt with it now. For just a little bit more out of pocket, many of the manufactures will sell you a recovery media kit for your system that frequently includes these diagnostics (Compaq comes to mind here). Trust me, it’s worth the purchase if you’re going to go this route of computing ownership.

On a personal note, I can’t stress or speak more highly of the benefits of running hot-swap hardware RAID-5 (or 6 is available) on your server. I can’t tell you the number of times I’ve had a drive fail and all I’ve had to do was walk over… pull the old drive, plop in a new one and bamm! Back in business, frequently without even having to reboot. The trick here though is having mixed age, mixed usage history drives so that you reduce the chance of simultaneous drive failures.

(note: this article was updated on July 14, 2010 for grammar and structural improvements)