By Joe Landman’s definition, when it’s an IT Cluster. Joe starts by quoting Doug Eadline from a recent article in Linux Journal, and it’s a good quote, so I’m going for an insideHPC first with the coveted “quote of a quote” followed by a triple axel. First Doug,
I would listen, then sigh, and explain that selling a rack of servers is not a cluster, the cake still needs icing. Viewed only in terms of hardware, the ticket to the HPC ball got extremely cheap. Even the rackem-stackem-fly-by-night-want-to-be-HPC vendor could get in the door. And yet, most of these systems are not operational clusters.
Then Joe, with some stuff that is generally acknowledged by HPC shops before they start a purchase, but then just as generally gets forgotten on the day that the purchasing decision is made and someone is staring at a 1.5x price difference and engaging in this kind of dialogue: “Let’s buy the cheaper one. It won’t work well and we’ll spend all our time trying to make it work. But it’s cheaper.”
I call these in general, IT clusters. They are not HPC clusters by any stretch of the imagination. They don’t really work well. Some things sorta-kinda work. Lots of things don’t or cannot. You have some interesting failure modes.
He then goes on with some “You might have an IT cluster if…” points to watch for. This is a good article that deserves your attention, especially if you have anything to do with buying hardware.
Joe and Doug, if we ever actually meet, this is my official offer to buy you guys a beer. I’ll be at SC this year, probably wearing a goofy insideHPC shirt.