Date: Tue, 04 Feb 2003 22:44:47 +0100 From: Poul-Henning Kamp <phk@freebsd.org> To: current@freebsd.org Subject: Preview: GEOMs statistics code. Message-ID: <25779.1044395087@critter.freebsd.dk>
next in thread | raw e-mail | index | archive | help
I have created a patch which contains a preview of the statistics code I intend to add to GEOM, and would invite people to play with it and give me some feedback: http://phk.freebsd.dk/patch/geom_iostat.patch Here is a small shell-script to watch the output with: http://phk.freebsd.dk/patch/geom_iostat.sh Here's the story: For each consumer and provider[0] in the GEOM mesh I collect three metrics: nop - Number of operations nerr - Number of non-zero error returned. dt - Accumulated request time. These three are collected for the three I/O operations: Read, Write and Delete[1]. (Currently the patch exports the values in ascii with a sysctl, this will change to binary once the prototyping has finished) If you run the shellscript above you will get auto-refreshing output like this (Apologies for wide lines): 0xc412db00 "ad0s4c" R 4160 0 0.002670 W 232389 0 0.006616 D 4083 0 0.001208 0xc412dd00 "" R 4162 0 0.002641 W 232390 0 0.005835 D 4083 0 0.000098 0xc412c800 "ad0" R 4188 0 0.002628 W 232390 0 0.005661 D 4083 0 0.000010 0xc412d500 "" R 4150 0 0.002655 W 232389 0 0.006616 D 4083 0 0.001208 0xc414e000 "ad0s4" R 4167 0 0.002646 W 232390 0 0.005835 D 4083 0 0.000098 0xc414e100 "" R 4168 0 0.002612 W 232390 0 0.005661 D 4083 0 0.000010 The colums are: Identifier - can be found in the "sysctl -b kern.geom.confxml" output) Name - Only for providers and only while prototyping. "R" - indicates read metrics follows. nop - Number of read operations nerr - Number of errored operations dt - Average respons time. (the, "W", nop, nerr, dt, "D", nop, nerr, dt) Now, discussion time: Collecting number of operations and number of errors is a nobrainer. The timestamps cost something to make, and my plan was to only collect them while a monitoring program is running. (Is this a good idea ?) In difference from the devstat framework which measures how big a percentage of the time a drive has one or more outstanding requests, I think that measuring the responstime is a much more useful metric. (comments, input, references welcome) I pressume we also want to collect number of bytes transferred, and I will add that in the next iteration. Comments, ideas, suggestions. Poul-Henning [0] A consumer is the "filehandle" a geom method uses to initiate I/O on. The BSD geom which splits ad0s1 int ad0s1[a-h] has a single consumer which is attached to ad0s1's provider. Consumers are unnamed and more than one consumer can be attached to a provider. A provider is a service point on a method and it has a name. In the above example, the BSD geom would have providers named ad0s1[a-h] [1] Delete is used to indicate sectors no longer in use in order to implement features like secure erase, pre-erase of flash devices etc. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 phk@FreeBSD.ORG | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?25779.1044395087>