This question as it stands is more or less meaningless, as are the rumored
values of 5-10%. There are two types of performance measures commonly used:
throughput -- the number of operations of a given type completed in a given
amount of time -- and response time -- the time it takes to complete a given
operation. To some extent, there are tradeoffs between the two. As load
increases, all other things being equal, throughput will increase and so
will response times for the operations. So I think the question you need to
be asking is, for any specific benchmark or application, "How does
throughput differ between the two environments with all other things -- load
and response time, for example -- being equal?" or "How does response time
differ between the two environments with all other things being equal?"
The other point I think needs to be made is that, while performance of the
*kernel* and its mechanisms is important to the folks on this list, it is
*not* for the most part what people buy computers for. For the most part,
people buy computers to accomplish *tasks*, for example running an
e-commerce business, processing astronomical images, editing documents,
creating electronic music, communicating with the Internet or operating a
manufacturing production line. How the *application code* performs, and *how
the kernel manages competing demands for resources from applications and
their users* are what matters. In other words, an inefficient kernel is a
bad thing, but inefficient applications on top of a perfect kernel are much
much worse.
-- M. Edward (Ed) Borasky http://www.aracnet.com/~znmeb mailto:znmeb@aracnet.comStand-Up Comedy: Because Man Does Not Live By Dread Alone
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun Sep 23 2001 - 21:00:25 EST