Partly due to flooding in Thailand pushing up disk prices, and no doubt partly due to the natural progression of new technologies driving down the prices of flash memory and related components, enterprise SSDs are now cheaper (at around $2 per GB) than enterprise 15K RPM HDDS (at around $3 per GB).
It’s different on the consumer side. Once you get into mainstream capacities, consumer HDDs are still under 20 cents per GB while consumer SSDs are typically $1.20-$1.50 per GB, depending on your comfort level with rebates. You can do better on both counts if you’re willing to watch Dealnews or Fatwallet and wait for a deal of course.
Enterprise HDDs have always been considerably more expensive than consumer drives, because they’re faster and more reliable than consumer drives. It’s been a couple of years now since I’ve worked directly on servers–I’m strictly a security guy now, during normal business hours–but when I did, the hard drives failed only 50% more often than memory failed.
Yes, that surprises me too.
On that job, I managed more than 200 servers, and in a typical year we lost about five of each. My rough estimate is that we had 1,200 memory modules in the field and around 800 HDDs. So if we lost five of each, our failure rate on memory was about .4% per year, and our failure rate on HDDs was about .625% per year.
It’s been longer since I’ve dealt with desktop PCs in a significant quantity, so my numbers are hazier. The last time I did that, I handled around 200 desktop systems, so that would have been a pool of around 400 memory modules and 200 HDDs. I remember memory failure being less than 1 percent per year, which would mean a maximum of 3 failed memory modules per year. Dead hard drives were more frequent, but not a monthly occurence. Six per year seems low, but it’s probably reasonable. That’s a failure rate of 3 percent.
That’s not a completely fair comparison, because those desktop PCs were replaced every 3 years, while the servers were replaced every 5 years, if we were lucky. And most of those desktop PCs were in the early stages of their lifecycle, while the servers were in the late stages of their lifecycle. But even when you compare old hardware with new, the server hardware was better.
And the failure rates on the server hardware were high enough that management finally authorized replacement of it around the time I left. The failure rates certainly made my job harder.
Smart companies buy reliable hardware and pay more for it. The reason comes down to pure survival. A business that goes offline for a week has a paltry 35 percent chance of surviving. (That’s a CISSP question, if you’re wondering.) It’s much cheaper to buy quality equipment and replace it before it becomes a liability.