-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Wed, 4 Jun 2003 15:00 pm, Andreas Schaufler wrote: > thank you very much for your answer. This measurings are done at my > university. We try to compare Windows 2000' against Linux' network > capabilities and are interested in getting meaningful results. Those shall > later be used in order to see which OS is to be prefered in a Parallel > Computing System from the network point of view. > Why do you think we won't achieve this goal this way ? What would be an > alternative ? Because the effects that you are trying to exclude from your measurements will almost certainly occur in the production environment. For example, does your parallel computing system do any scheduling? What makes you think that the networking differences will dominate the scheduling differences? What about quality of the toolchain? Or the VM performance? Even if we are just looking at networking, why do you think throughput is more important than (say) latency? Is the real system going to be point to point? What happens to the networking stack in the presence of collisions and retries? What happens if you need to change the networking hardware? As a more reliable guide to the best OS, look at what other users choose for parallel computing clusters based on commodity hardware, and figure out how your configuration, role and environment differ from theirs. Brad -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.7 (GNU/Linux) iD8DBQE+3aCeW6pHgIdAuOMRAnZtAJ0d76ah+K3D8QNbkl0skNNxQVSVuwCfWqtA /B/nHGQQwFgnJK7eBgnxb1I= =KYYe -----END PGP SIGNATURE----- - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
