Preamble: As part of the acquisition of a smaller development company we have acquired a new office. Many people were transferred to head office, but 'new' office will retain a small team of 4-6 developers and business analysts, along with other misc administration staff.

As part of that process we have ended up with several late 2005 model HP servers that are currently not being utilized which I feel could function as a suitable platform for UAT and source control, et. al., until next year when budgets are adjusted and a strong case for virtualization can be made.

The Problem: The existing sysadmin is technology gun-shy and spends most of his time focusing on job security (i.e. doing as little as possible) He has been resistant to my suggestions that upgrading these boxes would provide a minimally suitable infrastructure for the development team for the next year (I revised this value from 1-2 years based on feedback I've received).

The sysadmin's position is that we would be better off running things locally (UAT, etc.) on our desktops, instead of putting in the work to perform upgrades. This leaves the development team in limbo, with no guarantee of virtualization next year because the sysadmin lacks the experience and confidence to implement a solution.

This seems irrational in the face of a minimal, one-time, < 1k$ expenditure. The justification seems obvious to me as a short-term fix, but I want to be sure I am on base and not just desperate for a "fix".

The Question: What is your general process for calculating the cost/benefit for hardware expenditures, and how do you typically present this information to upper management to justify expenditures?

Additionally, how would you make a case for virtualization?

And, how much experience do you feel is required for a sysadmin to roll over to a VM environment?

  • I would break the "case for virtualization" questions into a separate post here on Server Fault. You'll get better answers and more attention by doing so.
    – ewwhite
    May 15 '12 at 22:28

Doing a cost-benefit analysis involves speaking the language of business, which is (mostly) money.

You have to demonstrate that the course you want to take is cost-effective. This can be done several ways, but you do need to break down certain costs. There are direct costs, indirect costs, as well harder to quantify things like opportunity costs and risks-to-the-business costs.

Some examples of the types of costs:

Direct costs

  • Cost of service-contracts of the servers. The older the server, the more expensive the contract.
  • Cost of keeping the servers powered and cooled. Older servers are less power-efficient than newer ones on a per-server basis.
  • Cost of parts replacement as things break (if no service contract).
  • Cost of keeping adequate backup protection.

Indirect costs

  • Cost of downtime per hour/day (in your case, it's probably lost-work time for the devs)
  • Cost of not having a service-contract.
  • Time charges for handling repairs by staff (this could be a direct cost, or it could be an indirect cost if exchange-time is used)

Opportunity costs

  • Development paths not able to be taken due to antiquated hardware (your stuff is old enough it might be on the wrong side of the 32-bit/64-bit divide)
  • Time lost due to emergency parts replacement
  • Time spent working around old stuff in order to get job done
  • Lack of virtualization makes existing work-flow less efficient

Risks to business costs

  • The likelihood of the indirect cost of downtime happening goes up with age

Quantifying all of these into money will go a long way to persuading the powers that be that you need more resources. It does take some good writing, knowing who you are talking to, and having an idea of the kinds of arguments that will convince them. Money convinces most management, which is why I'm suggesting it, but other approaches may be more valid; only you can tell that though.

Start with a single email and throw in some teaser numbers (with just-enough justification) to let them know you've thought about this, and only drop the stats-bomb on them when they ask for elaboration. You really want to avoid the tl;dr problem that most middle and upper management suffers from.


I keep it simple when presenting to the higher-ups. The hardware I deploy is meant to live a 3-5 year cycle, depending on application. 2005-era HP ProLiant systems have been revised FOUR times since then. That's four big jumps in processor, RAM, storage and power technology that should not be ignored.

The arguments are easy.

  • Technological improvements: Even the lowest-end system you can buy today is a huge step up from what you're using now.
  • Support: The present systems are way beyond end-of-life. Continuing to run them can have dire consequences. Finding parts and the resources to support them means scouring eBay. A good measure is checking the street value of your current equipment. If it's less than $200, you're on the wrong side of technology.
  • Virtualization: More effective use of hardware resources.
  • In a related vein to my original post - have you found virtualization difficult to push for, and if so, what steps did you take to handle it?
    – Ray
    May 15 '12 at 20:19
  • 2
    I describe virtualization as the only way to fully leverage the resources available in modern hardware. Utilization graphs help, but it's clear to people that most server workloads use far less resources than are available in the hardware. Virtualization helps consolidate and abstract the hardware into more usable chunks... Since terminology has changed, be sure to use the word "cloud" in your discussion :) Watch the heads nod in approval...
    – ewwhite
    May 15 '12 at 20:25
  • 1
    awwww you went and spoiled it ;)
    – user9517
    May 15 '12 at 20:26

I read three points in your story.

  1. I do not trust the sysadmin to make the right business choice.
  2. I got old hardware.
  3. I want to see if it is economical to use these.

I can not not anything about the first point. Your admin might have some good reasons which are not clear to you. Or maybe he has not. Personally I would **not* want to use 7 year old servers if I could avoid it. Seven years is old. Hardware will fail at some time and that will cost you downtime. How much you can afford that depends on your business. Is it more expensive for an entire office not to access a server than to buy new hardware now (rather than over 2-3 years)?

You write 'a small office', so I assume this point is less clear than it was in my case (single file server. 150 users all of which could not nothing if the file server failed). The price of a new server with equal performance to a 7 year old server compared to an extra day downtime for a dozen users might very well favour the new server. And that is with a new server which should last longer than 1-2 years.

Another point which might surprise you is the price for old legacy hardware. You can not just go to a shop and buy extra RAM for the same price as you the RAM in your current generation desktop (DDR). You probably need to shop around. < $1K seems optimistic.

Next there is support. A new server will usually get 3 or 5 years warranty and next business day supports for a price. Support for hardware older than 5 years tend to get very pricy. Often is is cheaper to buy a new server with 3year NBD support than to get a support contract for years over 5. This conflicts with your goal of doing this cheap.

Your could try to do this yourself. Which means multiple spare servers. And you should have a backup plan in case of failure which is more likely with old hardware. And unless the new office gets closed within a few years you will have to get a new server after a while anyway. Postponing that makes short term sense.


I started to answer your question.

Now, several edits later I got a lot of text, but all of it simply states: Do not use 7 year old hardware. It probably will not save money on the long run. Which means you can not justify the expenditure upper management.

  • You make some good points here about warranties, next day service and reliability considerations which I hadn't considered. The comment "you can not just go to a shop and buy extra same for the same price" is painting with an awfully broad brush. I would challenge you to justify this claim with some supporting links, or work to develop a response that is more specific to the question. thanks for the insights.
    – Ray
    May 15 '12 at 20:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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