On Thu, 30 Sep 2010, Luke S Crawford wrote: > da...@lang.hm writes: > >> I am in the process of doing this with a batch of systems purchased >> 5-6 >> years ago. > > > How long do you usually try to keep servers online? with my cost of hardware > (lower than usual, as far as I can tell, as I build) and my cost of > power (higher than usual, as I am in California) and my higher than > average cost to downtime, I'm usually looking to retire old servers > after about three years, but I don't know if that's the right policy. > > Do you usually do it based on the ratio of what you are paying for power > vs what you pay for a new server? or do you keep a batch of servers > online until you start seeing more than a certain amount of failures? > a mix of both?
I'm in a bit of an odd situation where power costs don't hit me directly, and I am a smallish part of a very large datacenter, so my power costs are in the noise anyway. In theory we are supposed to be replacing them every 3 years, but that's before the quarterly 'budget crisis' hits and they ask if there is any way to delay replacements. I've also been growing pretty continuously, so while in theory I have a 3 year cycle, in practice what I do is far more complicated. when new projects hit I purchase new servers. I don't go to the bleeding edge of performance and cost, but I try to go as far as I can to find the weet spot just before price starts skyrocketing (although, occasionally I just buy the best server I can because the project needs the performance) I then look and see if I have anyplace that looks like it's running out of steam, if so I put the new servers there and take the old servers and use them for the new project. With production, DR, QA, testing, Development, and my own lab, there's a lot of room to cascade things down. If I start seeing failures, then Ifirst makesure I replace that generation of equipment in production and then tell management that they won't be able to delay the purchase any longer The other push that I have for phasing out the old systems is when there is a technology change. The last two changes like this were the move to 64 bit system (where I was able to get rid of the lst 30-40 32 bit systems in the name of easing my workload by not having to maintain different kernels. I was running 64 bit kernels with 32 bit userspace so it wasn't complete system images to be maintained) and the move to Gig-E for everything instead of 100Mb (which eased cabling and eliminated a lot of driver headaches) I tend to go with whitebox systems, so as long as I have a vendor who will supply me with what I ask for, I don't run into many changes that cause compatibility problems. Other groups in the company go with the name-brand systems, and with direct head-to-head comparisons of uptimes, I am very comfortable with what I have been doing, while having faster systems and spending less than half as much as the other groups. David Lang _______________________________________________ Tech mailing list Tech@lopsa.org http://lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/