This is the mail archive of the systemtap@sourceware.org mailing list for the systemtap project.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
Other format: | [Raw text] |
Hello all, Based on the list's feedback about the previous script (particularly in relation to the "N/A" assumption), as well as a modified version from Dave Wright, I have rewritten this script as attached. Also included is a new 'probe end' block, inspired by 'entropo' from Serverfault (http://serverfault.com/questions/157612/is-there-a-way-to-get-cache-hit-miss-ratios-for-block-devices-in-linux). Note that this link is where I first started out from a couple weeks ago. This seems to offer much more repeatable and sane results, although it's not without fault. For instance, it's possible for an ioblock.request probe to register without a corresponding vfs.read.return hit. Hence, there's a check in here for cache_bytes < 0, which should (in theory) not be necessary. Fortunately, in my testing it appears to be a very small discrepancy, so I feel mostly safe about just pretending it didn't happen- it seems to be dwarfed by real I/O, and I probably wouldn't have noticed it if I wasn't testing on a totally-idle server. http://sourceware.org/systemtap/wiki/WSCacheHitRate is updated as well, including props for those who've helped out in some way. :) Thanks, Jake
Attachment:
cache-hit-rate.stp
Description: Binary data
Attachment:
cache-hit-rate.meta
Description: Binary data
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |