Re: A new scalability mathematical model better than Amdahl's law

From: aminer <aminer_at_videotron.ca>
Date: Sun, 19 Sep 2010 01:45:02 -0700 (PDT)
Message-ID: <af814964-a3f4-4711-b2b0-6c21bb2cdd29_at_q9g2000vbd.googlegroups.com>


Hello all,

I have thought more about my mathematical model

http://pages.videotron.com/aminer/scalability.htm

I have not take into account a factor, and there is still a problem with my mathematical model.

I explain:

When . in the worst case, all threads are waiting for the lock , my equation works.

For physical contention , my mathematical model works, and it does predict the retrogade throuput.

To model for example my lock-free ParallelQueue or lock-free Ringbuffer
or lock-free flqueue under contention, it works, and it does predict the
retrogade throuput.

But there still a problem with my mathematical model, if there is contention , and the scheduler schedules threads other than the ones that are waiting on the lock, my mathematical model doesn't work correctly .

But it was a good attempt from my part to try to find a general scalability equation that also predict the retrograde throuput.

Regards,
Amine Moulay Ramdane.
http://pages.videotron.com/aminer/ Received on Sun Sep 19 2010 - 10:45:02 CEST

Original text of this message