2009/8/4 עופר ברוך <[email protected]>: > Any thoughts? > Is my test case ok? Am I doing something wrong? > Is this normal behavior?
If you're able to collect raw monitor data from z/VM, that should tell us whether there were any issues outside Linux that can explain it. We're happy to look into the data for you. Send me a note off-list for details. For inside Linux, we'd need Linux metrics too. My experience with measuring disk I/O in Linux is that normally the large difference can be explained because one of the cases did not actually do I/O (or not as much as the other). For example, when you have enough memory compared to the data set (like your 300 MB in 1G) and the test runs shorter than 30 seconds, there is no I/O at all (at least not before the dd command completes). In that case you're doing a CPU measurement and your limithard may be impacting your results. Re: expecting great I/O performance: We don't make the disks spin faster. Your DASD subsystem is made up of simple consumer quality disk drives (well, 15K drives are not as bad as the 4800 drives you put in your netbook). Once you actually write to disk, that will be equally slow. Though you 300 MB/s will nicely go into NVS and be written out later... Mainframe I/O performance shines in that you do things in parallel, that does not mean that without things in parallel something will run faster (and empty bus does not drive 50 times as fast as one filled with passengers). Rob -- Rob van der Heij Velocity Software http://www.velocitysoftware.com/ ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
