[Spread-users] Performance Question

[Mike Perik]

First, my apologies, I had a bug in my timing
calculation.  I'm fixing that at the moment.

I hope to have better results by tomorrow.

I've included my conf file and related spmonitor
output for you inspection though.


Here is my conf file, I've reduced my Spread network
to two machines from which I'm running the tests.

xray is the broadcaster and frln03 is the client.

==========
Spread_Segment  225.0.1.1:5003 {
        frln03                  10.0.103.173
        xray                    10.0.103.121
}

DebugFlags = { PRINT EXIT MEMBERSHIP GROUPS }
=======

Here is a traceroute and ping of xray being done from
the frln03.
 


 1  xray (10.0.103.121)  0.104 ms  0.089 ms  0.090 ms
64 bytes from xray.foxriver.com (10.0.103.121):
icmp_seq=0 ttl=255 time=0.087 ms


Here is the spmonitor output, it is from the period I
ran a 10 minute test.

Monitor: send status query

============================
Status at frln03 V 3.17. 3 (state 1, gstate 1) after
2415 seconds :
Membership  :  2  procs in 1 segments, leader is
frln03
rounds   :  172915      tok_hurry :   38933     memb
change:       1
sent pack:      24      recv pack :   65639    
retrans    :      39
u retrans:      39      s retrans :       0     b
retrans  :       0
My_aru   :  160702      Aru       :  160702    
Highest seq:  160702
Sessions :       0      Groups    :       0     Window
    :      60
Deliver M:  160626      Deliver Pk:  160702     Pers
Window:      15
Delta Mes:       0      Delta Pack:       0     Delta
sec  :      14
==================================

Thanks for the help,
Mike


--- John Schultz <jschultz at spreadconcepts.com> wrote:

> Mike Perik wrote:
> > I've done another measurement.  For ten minutes I
> > publish data on 33 groups.  The server publishing
> the
> > data timestamps when it sends the message and when
> the
> > client receives the data it timestamps.  I diff
> the
> > two timestamps for each message and I'm now seeing
> an
> > average of >.6 sec delivery time.  This is still
> very
> > high.
> 
> As others have said, on a decent LAN network you
> should see delivery 
> latencies on a moderately loaded system of between
> sub-millisecond and a 
> few milliseconds.  The fact that you are not seeing
> this probably 
> indicates that (1) there is a problem with your
> configuration file (it 
> doesn't reflect the actual setup of your network but
> still manages to 
> work in fall back mode) and/or (2) there are faulty
> links between some 
> of the Spread daemons.  We can continue to try and
> help you with (1). 
> You need to run spmonitor and turn on status
> reporting for all the 
> daemons.  If you see any of the retrans fields are
> increasing 
> significantly (in the hundreds or thousands), then
> this indicates that 
> your problem is either 1 or 2 and we can help you
> figure out which 
> machines are causing the problems.
> 
> Also, if you are willing to post your test programs
> to the list, we can 
> run them on our networks and see what kind of
> performance we get.
> 
> -- 
> John Lane Schultz
> Spread Concepts LLC
> Phn: 443 838 2200
> 
> 
> _______________________________________________
> Spread-users mailing list
> Spread-users at lists.spread.org
>
http://lists.spread.org/mailman/listinfo/spread-users
> 



		
__________________________________ 
Do you Yahoo!? 
The all-new My Yahoo! - What will yours do?
http://my.yahoo.com