small server: build or buy?

[CaptainPinko]

I'm working on a project that is building a website for my school's daily student newspaper (it is a large university). Part of the project is writing a proposal for purchasing a computer to server that page. Me and my partner have been wondering if it is worth building a server from scratch or whether we should be looking at some SMB entry-level server from HP or Sun or something.

The server will be doing apache and some basic MySQL\PostgreSQL (searching for text in an article, the article is all stored in a field of type "note" in MS Access currently and the database is only ~9mb currently). We plan on running it on Linux/OpenBSD/FreeBSD. The only requirement is that the server have at least RAID 1.

The concern was that after the effort and cost of individually selecting quality-parts to build the box might not out weigh the cost for just buying a server. While 90%+ uptime is not critical, after we are done the server will be ignored and sitting in a corner so we can't assume that any basic problem will be solved (such as RAM module dying but server would run if someone remove the faulty stick).

[GnomeProject]

I just finished searching E-Bay for lightly used dual and quad processor servers with Raid configurations. Considering I have bought big ticket items on there before, I don't hesitate recommending it, but I do suggest "BUYER BEWARE" if you go that route. You can find a decent dual 1ghz xeon or PIII server with 60 gigs of SCSI storage in a raid configuration for purty cheap (ie. less than 500). So maybe go that route? If not, new parts are gonna cost you a lot to pursue a "newer" version of these servers. But, if you go with a new or manufacturer refurb from a well-known company you also have the added benefit of a warranty...just something to chew on. Hope that helps?

-Adam

[heckle]

You could always try over at retrobox...

http://retrobox.com/rbwww/home/

[Hmse]

Well, if it's going to be a server, than you don't need that fast of a computer, so you could get individual parts for pretty cheap, but then you could also just buy some old box off someone.
Either one is a good way to from my eyes.

[EnigmaOne]

Given that it's a "large university", I'm willing to bet that you're underestimating your immediate and future traffic on the box.

What kind of traffic are you anticipating for the server?
Have you studied the matter enough to be able to quantify your traffic expectations?
If so, experience--in such a setting--says to double it.

The problem with a pre-rolled Sun/HP box is when something does fail, it's going to be far enough down the road that you'll probably have to replace the box outright because of parts unavailability.

You can roll-your-own, and probably come up with a total cost far below that of a proprietary server chassis, and benefit (in terms of longevity) from best-of-breed parts selection.

[soulestream]

yeah how much traffic? we have set up linux/unix boxes with 50 clients doing file serving/light webserving on PII 400 with 512 ram. pretty lightweight.


soule

[CaptainPinko]

well right now traffic is neglible... hence why we were called on to revamp the site. The university has about 30k ppl. I'm guessing 15k are undergrads whuch is the main audience of the paper. So if 2/3 read the paper thars 10k readers. Assuming about 1/5 to 1/10 will view the site instead of grabbing the paper that would be 1.2 hits a day... seems a little high but once this gets done we plan on pushing the site in the paper paper so not unreasonable. 'm gessing the hits while be grouped fairly tight together. The university has also been growing over the past 10 years so I expect the number to rise and since most of te campus is getting wifi I'd expect increase in traffic.