Samba and Application Response Measurement (ARM 4.0)

David Collier-Brown davecb at sun.com
Mon Mar 10 22:30:14 GMT 2008


  You're starting with some fairly hard stuff.  It's
easier to find the callers of (sys_)read, for example,
in recvfile.

  For a simple server transaction, it starts with a read
of the socket, which gets a request from the client.
The first write to the same socket ends the latency
period, and the last write before a read ends the transfer.
period.  The time between the last write and the next read
the think time of the client user of system.

  Once you've done the really coarse measurement, you can
drill down to the detailed requests, which will begin
with the decoding of a request packet (or and_x) and
end with the last reply.

  Someplace down there you're eventually going to be doing the
authentication and mount code...

--dave

Sven Kubiak wrote:
> Is there any technical documentation where i can find which part of the
> source code (or function) is responsible for:
> 
> - user authentication (to PDC)
> - loading the user profile
> - mounting paths and printers
> 
> As these are the parts we are most interested in measuring. Looking at the
> source code, i can see the different files involved in the mentioned
> processes, but it is difficult to pinpoint the right locations for the ARM
> transactions (i.e. the right functions).
> 
> Best regards
> Sven
> 
> On 04.03.08 20:28, "David Collier-Brown" wrote:
> 
> 
>>Volker Lendecke wrote:
>>
>>>On Tue, Mar 04, 2008 at 04:34:51PM +0100, Sven Kubiak wrote:
>>>
>>>
>>>>Implementing an ARM transaction to the main code was just to see if and how
>>>>ARM and samba would work together. Opening a file via network and measuring
>>>>the reponse time seemed practical for the first step. However, we are also
>>>>interested in measuring authentication, print jobs and other important
>>>>measuring points in samba.
>>>
>>>
>>>Is just wrapping fd_open sufficient for your needs? You
>>>*might* also want to measure the delays introduced by oplock
>>>breaks and/or share mode violations.
>>
>>Normal practice is to wrap a transaction, so you start the timing
>>when you get a fie read request, record the time of the first reply
>>and close the transaction when you send the end of file.  Similarly
>>for write, oplock break, etc.
>>
>>--dave
>>--
>>David Collier-Brown            | Always do right. This will gratify
>>Sun Microsystems, Toronto      | some people and astonish the rest
>>davecb at sun.com                 |                      -- Mark Twain
>>(905) 943-1983, cell: (647) 833-9377, home off: (416) 223-5943
>>(800) 555-9786 x56583, bridge: (877) 385-4099 code: 506 9191#
> 
>>
> 

-- 
David Collier-Brown            | Always do right. This will gratify
Sun Microsystems, Toronto      | some people and astonish the rest
davecb at sun.com                 |                      -- Mark Twain
(905) 943-1983, cell: (647) 833-9377, home off: (416) 223-5943 
(800) 555-9786 x56583, bridge: (877) 385-4099 code: 506 9191#


More information about the samba-technical mailing list