CIFSSTAT

[David Collier-Brown]

"Christopher R. Hertel" wrote:
> 
> >       I have a LD_PRELOAD hack to capture performance
> >       stats, what other things are we interested in?
> 
> Kewl.  What stats does it capture?

	Transfer times for files, plus overall rusage stats.
	The following is a run of a trivial test program that
	doesn't actually move any files:
--
yyyy/MM/dd:hh:mm:ss fd user file  size latency xfertime total bytes/s
2000/03/11:20:24:56 3 0 <socket> 0 -1.000000 -1.000000 0.000463
0.000000
2000/03/11:20:24:56 3 57957 /dev/udp 0 -1.000000 -1.000000 0.000249
0.000000
2000/03/11:20:24:56 4 0 <socket> 760 0.260443 20.000166 20.260609
37.511213
2000/03/11:20:25:16 4 0 <socket> 760 0.257572 19.999866 20.257438
37.517084

struct rusage {
  struct timeval clocktime = 40.529 /* clock time used */
  struct timeval  ru_utime = 0.10; /* user time used */
  struct timeval  ru_stime = 0.10; /* system time used */ 
  long  ru_maxrss = 0; /* maximum resident set size */
  long  ru_idrss = 0; /* integral resident set size */
  long  ru_minflt = 0; /* page faults not requiring physical I/O */
  long  ru_majflt = 0xd; /* page faults requiring physical I/O */
  long  ru_nswap = 0; /* swaps */
  long  ru_inblock = 0xd; /* block input operations */
  long  ru_oublock = 0; /* block output operations */
  long  ru_msgsnd = 0; /* messages sent */
  long  ru_msgrcv = 0; /* messages received */
  long  ru_nsignals = 0; /* signals received */
  long  ru_nvcsw = 0x2d; /* voluntary context switches */
  long  ru_nivcsw = 0x4; /* involuntary context switches */

	It's a framework for getting per-file latency, transfer
	time, and speed, and **could** be adapted to logging
	perormance stats inexpensively.

--dave
-- 
David Collier-Brown,  | Always do right. This will gratify some people
185 Ellerslie Ave.,   | and astonish the rest.        -- Mark Twain
Willowdale, Ontario   | //www.oreilly.com/catalog/samba/author.html
Work: (905) 415-2849 Home: (416) 223-8968 Email: davecb at canada.sun.com