The question is not "How much does it cost to setup?" It's "How much does it cost to be down?" Buck -----Original Message----- From: shrike-list-admin@xxxxxxxxxx [mailto:shrike-list-admin@xxxxxxxxxx] On Behalf Of John Haxby Sent: Monday, September 22, 2003 3:40 PM To: shrike-list@xxxxxxxxxx Subject: Re: [OT] The right HW for the each SW server Rodolfo J. Paiz wrote: > > 3Ware has been around for quite a long time and has excellent > support... I would definitely already consider them "conventional" > technology. As for SCSI vs. EIDE vs. SATA, I'm sure they all work and > don't by themselves affect MTBF in any meaningful way. I agree. My point is that you need to buy well-supported hardware. If you can look after the hardware issues yourself and fix servers when they go wrong, go with whatever you fancy. We have two lots of servers in our rack -- a bunch of Dell 1U machines and a bunch of cheap DNUK servers. The Dell boxes look after the critical stuff -- external DNS, mail, web and whatnot. The DNUK boxes look after internal things, software building, NIS, etc. If one of the Dell boxes goes wrong, we call out Dell and it gets fixed within a few hours. If one of the DNUK boxes goes wrong, we swap around hardware to get the most critical (at the time) service working again and the broken machine gets repaired in a couple of days by us. So, do you have the resources (I assume you have the skill) to maintain your own servers? Or would you rather pay for someone else to do it? It's certainly cheaper to pay for the extra to have machines maintained by someone else than it to take someone else's time -- and it will always happen at the least convenient moment. So if you're after a maintained solution, then you need to go with whatever your favourite supplier will supply. I checked Dell -- so far as I can tell the PowerEdge servers are SCSI once you get above a couple of internal IDE controllers. I think it boils down to this: 1. If you can look after the hardware yourself then you'll be able to make your own decisions based on whatever you feel comfortable with. If you're skilled and you have the time, you can get the best possible repair time here -- provided you maintain a complete set of (working) spares. 2. If you can't look after the hardware, pick a supplier that will provide good on-site service. You can pay for a given SLA which allows you to plan for the worst knowing what the worst is. Of course, there are plenty of small businesses (and quite a few large ones) that just buy whatever is available and trust it to be reliable. And almost all of them never have any problems. But there are thousands of companies that fold every year(*) because of some unplanned for hardware problem. jch (*) Not all of them fold at the point at which the hardware failed. Some hang on for as much as a couple of years, but the failure can be traced back to the original hardware problems. It's a question of what level risk you are prepared to accept. > > -- Shrike-list mailing list Shrike-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/shrike-list -- Shrike-list mailing list Shrike-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/shrike-list
