views:

156

answers:

1

I am thinking about getting one of these (PCI) to set up an internal entropy pool similar to this service who incidentally brought us fun captcha challenges.

Prior to lightening my wallet, I'm hoping to gather feedback from people who may be using this device. As there is no possible 'correct' answer, I am making this CW and tagging it as subjective.

I'm undertaking a project to help write Monte Carlo simulations for a non profit that distributes mosquito nets in Malaria stricken areas. The idea is to model areas to determine the best place to distribute mosquito nets. During development, I expect to consume gigs if not more of the RNG output. We really need our own source.

Is this device reliable? Does it have to be re-started often? Is its bandwidth really as advertised? It passes all tests, as far as randomness goes (i.e. NIST/DIEHARD). What I don't want is something in deadlock due to some ioctl in disk sleep that does nothing but radiate heat.

This is not a spamvertisement, I'm helping out of pocket and I really, really want to know if such a large purchase will bear fruit. I can't afford to build a HRNG based on radioactive decay, this looks like the next best thing.

Any comments are appreciated. I will earn zero rep for this, please do not vote to close. This is no different than questions regarding the utilization of some branded GPU for some odd purpose.

Answers pointing to other solutions will be gladly accepted, I'm not married to this idea.

+2  A: 

First of all, I can't comment on the reliability device you mention.

In general, Monte Carlo methods do not require the same randomness as cryptography does, so I think getting one of those might be a bit overkill. But it would probably work like a charm;-)

For your application, go with the Mersenne Twister which is pretty standard for MC.

Good luck with the project!

trolle3000