Architects Zone - Comments for "Build Multi-core Ready Java Applications, Part 2"

Great article, it was a

aliao — Wed, 16 Jul 2008 02:33:48 -0400

Great article, it was a bridge for me to explore more on DataRush, it's been an exciting subject to search on how to exploit capabilities that mult-core hardware can offer. Thank you very much for putting effort on writing the article in the first place, please keep up with good work.

Thanks for the comments.

ericjbruno — Tue, 29 Apr 2008 19:56:43 -0400

Thanks for the comments. This isn't my product, so I'm certainly not pimping it. I'm only writing on my experiences using it. I can't go into detail about all of the projects I've worked with using DataRush, so the numbers I've published are for the sample only. But real-world examples show that it does make magnitudes of difference.

You're correct, I did make it seem a bit trivial when there was some XML that had to be written (as opposed to Java), but the point was that I didn't have to write multi-threaded code that carefully synchronized access to shared resources, and I didn't have to account for exactly how many cores would be available in production; that was all done for me. That was the point. As servers are introduced that are 64, 128, and 256-thread capable, such as those by Sun, and so on, frameworks like this deserve to be looked at.

Again, I'm sorry if you feel mislead in any way. The first part of the article was purposely devoid of any product-specifics to try to avoid this.

This might or might not be

mossprescott — Tue, 29 Apr 2008 19:38:29 -0400

This might or might not be an interesting product, and I'm certainly not saying your title doesn't refer to an important topic, but help us out here:

Make it clear that you're basically pimping a product. Ideally, call your article "Build Multi-core-ready Java Apps with Pervasive DataRush."
Tell us something about the product up front (is it open source, who's developing it and for what perceived market), so we'll care.
Don't make claims that are obviously silly; 3 seconds isn't "magnitudes faster" than 8 seconds, it's a little more than twice as good. And you only scaled to multiple cores "without writing any additional code" after you refactored your entire app to be structured as a tree of operators (or whatever). That hardly seems like a trivial step.
If there really is some magic making the dual-core version more than twice as fast, please enlighten us, because it isn't obvious and sounds bogus without explanation.

I want to give you the benfit of the doubt here, but as is, this article doesn't even make me want to google the product (thus proving that I'd rather write a grumpy comment than learn something :)