Drinking the koolaid, by the megaliter

We have been talking about application acceleration, and heterogenous computing for quite a while now. Call it HPC and you scare off anyone who might otherwise be interested in helping to build the future of computing. It really doesn’t matter what you call it at the end of the day. It is coming. Fast. Fine. Lets call it Accelerated Computing (AC for short). I have a reason for this.

Accelerated computing is not just for high performance computing. Its for anyone with a project that grows bigger, requiring you run more things, with higher accuracy, quicker than before. In other words, almost anyone doing any sort of serious computing. Accelerated computing is not just HPC, rather HPC has been an early adopter of acceleration technologies. We now (as in today) have prospective customers for acceleration technologies. Outside of traditional HPC. What is interesting is that this customer asked us about this. We didn’t have to pitch them. We didn’t need to mix the koolaid. They came in being believers. Their problem is definitely amenable to accelerated computing. Another group, a newly formed entity that I am aware of and have talked with one of the founders of, is looking to use accelerated computing technology to gain a significant competitive edge in their field. I won’t spill their beans; heck, I would like to help them succeed. If you can deliver a solution, faster, and lower cost than previous groups, you may have a chance at a larger market, provided your faster/lower cost solution solves a problem that needs solving. Computing infrastructure ages, and as computational workloads increase, this aging represents an increasing risk to end users in terms of getting their work done. So groups with ever-increasing data and computational needs will be continuously looking for items that address their pain points. Along comes an accelerated computing solution. The technologist in me really wants to delve into the inner workings. The business person in my asks “what compelling need does this solve?”. The answer to this is quite simple. If you can decrease your cycle time from 1 per week to 4-5 per day, you can change the way you work, ask better questions of your model, and refine it further. Your productivity will increase. Not sure if this really works? Go ask a few engineers and scientists who leverage clusters for their own computing whether or not they like waiting a week or more in the national level queues, or prefer the immediacy of their own calculations on their smaller local personal supercomputer. Now there is a marketing term. Last year a person at a large OEM was excited when he said he could deliver a teraflop in 3 racks. I asked him if he would prefer a teraflop in a 4U package. Rapid blinking.