Re: [benchmark] Web applications benchmarking

Alexander, The mails have not been delivered to me for some reasons. Thank you for forwarding the mail and sorry if I missed something in the context. > Would this scenario be a good setup for a useful web application benchmark? > The script automatically sets up three VM's: one for webserving, one for database and one > for the client. > The VM's make sure they can connect to each other. > The Client VM start accessing the webserver and logs the amount of requests per seconds. > Rubis have some documentation of their system with 3 videos that you might take a look at. It sounds reasonable to measure DSL (PHP) runtimes in a virtualized environment, as the virtualization does not likely affect the performance of core language feature execution, at least according to our private experiments comparing PHP performance on Xen and native environment. You may want to prove that the measurement in your pseudo environment, not only with virtualized host but possibly also with other "fake" environmental setup, is equivalent to, or has a tight correlation with, the one in the real enviornment. Regards, Mich - Michiaki (Mich) Tatsubori, PhD. mich@acm.org IBM Research, Tokyo Research Laboratory