0

I use an application that does CPU intensive multithreaded optimization which is currently run on a hyper-threaded Quad Core machine that it just pegs at 95-100% utilization across all cores for hours. I am planning on running it on a 32 vCPU EC2 instance, but even then it could probably benefit from additional processing power.

My question is in regards to creating a cluster or similar to run this type of program on. This is a 3rd party 64bit Windows application that wasn't necessarily designed for distributed computing, and I am having a hard time figuring out if it is even possible to cluster Windows Server 2012 for this purpose.

From what I can tell I need something similar SSI (single system image) cluster that would allow the application to run like it is just a single powerful computer.

In a Windows environment is something like this or similar possible?

user3033725
  • 3
  • 1

1 Answers1

0

No, not possible. SSI is not supported outside of the hypervisors doing it. If you get one and use it then you can run windows on it, but windows by itself has no SSI capabilities.

TomTom
  • 51,649
  • 7
  • 54
  • 136
  • So what your saying is in theory I could create an SSI and then run Windows on top of such a system, but other wise there is no other good way of going about this on a Windows system? – user3033725 Nov 25 '13 at 20:55
  • Exactly. No SSI support IN windows - running it as client does not require OS support. That said... this is the reason I rewrote our inhouse main software to be inhouse / not purchased... missing grid-capabilities. – TomTom Nov 25 '13 at 21:08
  • In theory from a technical stand point how hard is it to make a multithreaded program cluster computing capable? – user3033725 Nov 25 '13 at 21:31
  • Depends totally. In my case we did a total rework. We also moved from "program controlled" to database controlled so that in case of a restart the work continues where it was last recorded in the database. Was worth it for us - this is our core business application and we are now maybe 1000 times faster than before and scalable. But generally it is a VERY Large undertaking. – TomTom Nov 26 '13 at 07:19