How do you guys handle obsolete servers? When the "world" expects to be running on 2.5ghz PVIIs, and you are running 650mhz PIIIs, will you simply move all your existing clients to the new hardware, or not do anything till the old hardware dies?

How often do you plan to roll your servers out of your NOC or Colo facililty?

Do you capitalize your servers? Do you not because you replace them in only 3 years anyway?

I had one PM reply to this post, and that said that no one has replied because no one obsoletes servers.

Is that true? Do you count on customer turnover and just keep running until when - until you have a catastrophic hardware failure?

I can't believe this is the state of the industry. Some of you must have obsolescence policies.

-t

Our obsolescence policy is simple:

Is the server adequate for the task at hand?

If the server was adequate when it was purchased, it does not cease being adequate for the same task a year or two later. Just because newer machines are faster does not mean that the older machines have gotten slower over time.

This isn't like a vehicle where performance degrades over time.

So Travis, what do you do when the hardware ages to the point where hardware support is difficult?

If you have a hardware failure on a leased server, do you replace the server if repair costs are going to run 50% of a new box?

If you lose the motherboard on a 600mhz Intel box in two years, what will you replace it with?

Are all your servers rack mount and built from scratch? Can you simply plug in someone elses motherboard and go along your merry way? What about SCSI drives? If you have migrated to another drive configuration for all your new servers, will you stock replacement drives for your older technologies as well?

I know this is all based in speculation, but planing for the future and what may happen when, is always a good thing to do.

-t

We've been fortunate, actually - the only hardware failure we've ever had has been disks. (And those are always the number one failure point.) Whenever we purchase a particular type of hardware, we inventory spare parts, so availability down the road isn't really an issue.

In practice, we usually look for about a 2 year service life out of a server. We've been in business three years now, and the first servers were retired out a little over a year ago. This was done because we needed higher performance out of them, though.

The current majority of our servers were built in-house. As such, they have completely interchangeable hardware, which is nice. Let's say a motherboard fails, as in the hypothetical case you mention. Even if a replacement isn't available - no big deal. A current motherboard/processor combination can just be swapped in. PC parts are standardized enough these days that motherboards and processors are the only parts that really obsolete quickly.

These days, we're buying Dell systems, which use custom motherboards. We deal with this situation by just buying spare servers. Except in the case of something simple like drive failure, we just treat the entire machine as a FRU and swap in a good server in case of trouble.

I agree that it's good to plan for the future... and good basic policy like buying spares for equipment along with that equipment will always serve well. (That's good sense that applies to any kind of hardware in any kind of industry.)