Better resource management for “old” apps

I was building an OCR driven application, implementation of EMC² Captiva InputAccel as input management system for enterprises. We had a server dedicated just for OCR mechanism as we were having a higher volume of scanning/documents to OCR on a daily basis.

OCR mechanism was an “old” application – an Abbyy FineReader OCR module for InputAccel, written obviously before multi-core CPUs etc, but it was good enough to do the job.
This app is capable of running in a multiple instances as in that way it could complete it’s processing faster. But, there was a problem.

No matter how fast server we built (hardware) it would always use 100% of CPU when we try to run only two instances – and processing wasn’t faster. Yes, I know, you would say – but OCR is CPU intensive processing – yes it it, but shouldn’t it perform faster with a much faster server then the one we initialy buil? Well, it didn’t.

So I came to idea to try with virtualization. I installed a free virtualization engine – Microsoft Virtual PC 2007, run two intances of Windows XP on it, and run two instances of OCR app on each of them – and voila ! That solved the problem with ease.
This was a good start for us with this issue as we now have confirmed that this OCR app works in a virtual environment without problems.

This free MS VirtualPC is not capable of dedicating custom number of CPUs/cores per virtual machine, it by default dedicates only one CPU, but that was good enough for us as we even intended to dedicate one CPU per machine anyway.

So, as we are running high volume scanning/processing, for a production environment we however wanted someting more stable and decided to go with VMware, as we have more freedom in it’s configuration for fine-tuning then with MS Virtual PC.