On Thu, Nov 20, 2008 at 11:53, Andy Michaels <[EMAIL PROTECTED]> wrote: > Hi everyone. Over the years, there has been much talk about network > performance of various Soekris models. I think it might be > interesting to post some standardized results based on agreed-upon > metrics and test cases.
Metrics & test cases will probably be easy to agree on; what constitutes appropriate tuning on the candidate OS and whether to allow those tunings will likely be harder. I personally would like to see either multiple untuned OSes or a single carefully-tuned one generally accepted to have the fastest stack. If this is to go on for a while you could add the tunings at a later point, but as is usual with such efforts, the first round is usually the only one that gets done. > On the software side, we'd need to agree on a set of OSes and configs > as well as defining a test suite. I can stub out some sample test > cases on the wiki, but I'll need the community to complete the list. Stub away. Mousetrap and all. > Also, please let me know if you think this is a worthy use of time and > resources, or if you think it's a complete waste of time. Benchmarking is generally less worthwhile since it seldom reflects real-world scenarios. Unless you have a huge amount of time on your hands, I would suggest aiming at edge cases (worst & best scenarios) that outline the performance envelope instead of trying to color it in. _______________________________________________ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech