Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm trying to estimate ROI for a new server and how long it will most likely be in service. Are there any average numbers for how long you can expect a server to last that's in 24/7 use?

In other words, is it likely to find 10-year old server from 2010 in today's racks? How about servers from 2005? I'd love to get an idea for how long a server will most likely be in use.

By the way, this is meant irrespective of growing or changing application needs. I'm talking only about the hardware.


locked by HopelessN00b Jan 22 '15 at 3:17

This question exists because it has historical significance, but it is not considered a good, on-topic question for this site, so please do not use it as evidence that you can ask similar questions here. This question and its answers are frozen and cannot be changed. More info: help center.

closed as off-topic by HopelessN00b Jan 22 '15 at 3:17

  • This question does not appear to be about server, networking, or related infrastructure administration within the scope defined in the help center.
If this question can be reworded to fit the rules in the help center, please edit the question.

possible duplicate of Whats the average lifespan of a server? – RobM Dec 8 '10 at 18:56
This question is off-topic under current topicality rules. – HopelessN00b Jan 22 '15 at 3:17
up vote 11 down vote accepted

Servers are generally built with fairly high quality components, especially from big brand names. In my experience they can easily run for 5 years easily before becoming unreliable. (replacing the disks in the raid array along the way).

However, they may not be economical in that respect. The capacity/power of new machines keeps rising dramatically, while your overhead costs (human and power) remain expensive. At some point the raw capacity becomes just not worth it.

+1 for not focusing on the hardware. A decent server will ither fail fast (bad hardware) or last way longer than 5 years, ignoring discs which are hot swappable. The main reason really IS the age capacity. I recently replaced 3 machines 4 ýears olf with one rnuning virtualization - and that one still now has 24gb free memory (32 gb instead of 12). The power / space savings pretty much pay for the hardware. – TomTom Dec 8 '10 at 8:51
That matches with my experience, although while I expect an average usable life of 5 years, for accounting purposes I use a figure of 4 years to be obn the safe side. – symcbean Dec 8 '10 at 10:48
same here. i DO have 6-7 y/o machines running; but usually consider them to be disposable after 3-4 years. If it's going to be dedicated to a client that asks about 'his' hardware, i offer hardware replacement every 3 years. – Javier Dec 8 '10 at 11:19
+1, for reasons already listed. Most server we replace every 3 to 4 years (same for desktops and notebooks). I've got one at home though that's 9 years old and still kicking; had one 12 years old that just died (too hard to find replacement parts). – Chris S Dec 8 '10 at 18:38
Bigger data centers turn the machines off once they're of the lease contract (most have an replacement plan or when the financial write-off of the equipment is expired. – pfo Dec 13 '10 at 11:58

Its not likely to find a 10 year old server in the racks unless it has been re-purposed, or isn't replaceable. If nothing else it becomes difficult to find replacement parts. For production use I usually schedule replacement every five years or so. If the budget doesn't allow for purchase of development hardware, they may end up being recycled as development machines.

Server grade hardware is likely to run more or less forever except for moving parts. Support, power requirements, and physical size to performance become the significant factors.

I believe our tax rules allow for a full write-off over three years (actually four reporting years). From a technical standpoint anything you can buy is likely already obsolete. From an accounting standpoint, if it still does the job it isn't obsolete. The performance/cost ratio is continuing to fall and may make earlier replacement a good option.


Generally, if you're buying a server from one of the big vendors the service-contract for it will be pretty cheap until it turns 5 years old. At that point, if you can even get one, it'll be VERY expensive. For our hardware vendor the 1 year service-contract cost at Year 6 is about 50% the cost of a new server. This is how they encourage regular hardware replacement. You are factoring in the service-contract in your ROI calculations, right?

Once a server turns 5 around here it tends to get relegated to test/dev/low-impact roles and we keep a grave-yard of similar machines for parts if it comes to it. Those of us who keep this hardware running frown very discouragingly when entities ask to use old stuff that we have 'just laying around'.

My oldest monster right now is an ancient HP LH3, a 12 year old server, that is performing a single task: monitoring our datacenter UPS internals. The next oldest is a 9 year old that's due to be replaced in the next 3 months. Neither of these have service-contracts, and if they die we will sigh deeply but won't be otherwise inconvenienced.


This can't be said as confirm it is totally depends on how servers has been used and what was the use of server. What application installed on server. How much load was server having.

All these calulation matters regarding server performance.


The answers to date are good, I just like sharing stories:

I have seen servers last over 10 years. When I was at a $BIGPHARMA in 2005, I performed decommissions and retirements in the datacenters. I retired a handful of old HPUX cabinets, did some SGI IRIX machines (Indys, I think), some Netware boxes that had to be near 10 years, and at least two Sun SPARCstations. One pizza box, one lunchbox, and I think they were both running DNS servers.


50,000 Power On Hours is a magic number. That works out to about 5.7 years. The number of issues we see with solid state components at and beyond that point is drastically higher regardless of the kind of server.


We used to plan financially to depreciate machines after 3 years. At which point they were classed as essentially worthless.

We used to plan to run the hardware for about 4 years, and the machines were often decommissioned at around 5 years old. At this point they were moved to non-production.

At 3-4 years, the machines were well used. We'd buy new kit, but until it had been running for 6 months we wouldn't dream of moving them into production.



Not the answer you're looking for? Browse other questions tagged or ask your own question.