> The one thing that always prevents me from using SCSI is replacement time. > If the hardware fails, there are a million little PC shops I can go get a > SATA drive at; SCSI is an order-in event, even in most bigger shops.
I've tried to keep out of this convo as I have some strong feelings about it, given that I've faced the issue(s) over and over again. These days I have a simple strategy - Assuming you are a small shop: Step 1: Buy big brand-name server with multiple dual core procs, vmware ESX, gobs of memory(~24GB), and lots of scsi disks in HW RAID10 (dell 2900 is probably good bet here, and supports 10 hotswap SAS drives) Step 2: Get a vendor gold support contract with 2 or 4hr guarenteed response time (that includes replacing all disks that fail over a 3-4yr period). Step 3: This is the one and only server you have for the next several years, sell all your old ones on ebay. Assuming some minimal vmware support and other costs, your amortized bill should be about $5K/yr total. That is a very reasonable hardware + support budget for a small company. Step 4: Congratulations, you have relatively good reliability and response time given a small-medium budget, your ongoing maintenance and management costs are negligible, and you have substantial ability to bring online new virtual systems (as many as 24) within minutes and/or to micromanage allocation and prioritization of cpu, memory, and disk resources. For basic backups, you can take snapshots of virtual machines and send them off to tape or some other cheap SATA disks. No need to keep spare drives around. Once again, this is your one and only server box and the power/rack space savings are also nice. MattM -- Matthew Marlowe ([EMAIL PROTECTED]) Yahoo IM: deploylinuxconsulting -- [email protected] mailing list
