I could pat my own back on this … no really, I could. Wouldn’t be hard.
I’ve been talking for a long time about how the HPC market will likely evolve. Hidden within this is how to grow as a business … serving this need.
We’ve been predicting that the cloud HPC model will reduce the number of new clusters deployed. Basically, acquisition costs for running a cluster are large, as well as the lifetime costs. In this era of declining budgets, and budgets under pressure, people are looking to reduce this expenditure. If you can drive your acquisition cost to zero, and simply pay for what you use as a service, then you may be able to reduce your upfront costs, and longer term manage your cost structure.
Moreover, we’ve been noting that we see a muscular desktop as a growth market. These aren’t the typical little Dell or HP desktop boxes. These are beefy units with a great deal of processing power, ram, storage and storage bandwidth. Our machines today can have up to 32 Intel and 48 AMD processor cores, with 288 GB on Intel, and 512 GB on AMD, multiple GPUs and a more than 1GB/s IO channel, right next to your desk in an almost silent chassis.
This local capability, combined with its far lower acquisition and management cost as compared to a cluster, suggests that this will a preferred direction for wide spread HPC. The article suggests that people see this happening now.
Yeah, thats about what we expected to happen. More local storage needed, more high speed storage local and remote, and more local desktops and cloud HPC machines. The only impediment to this appears to be networking. T3’s are slow and expensive. Gigabit to the premises is very expensive. Well, here in the US it is.
But I think this is changing.