[Spread-users] Re: Spread Performance Questions

[Daniel Marques]

Interesting: As the message size grow to 8KB, the disparity reduces
dramatically:

$ ./test.exe C
UNRELIABLE_MESS: elapsed time is 4 second 375000 miliseconds
RELIABLE_MESS: elapsed time is 4 second 391000 miliseconds
FIFO_MESS: elapsed time is 4 second 312000 miliseconds
CAUSAL_MESS: elapsed time is 4 second 422000 miliseconds
AGREED_MESS: elapsed time is 4 second 406000 miliseconds
SAFE_MESS : elapsed time is 6 second 188000 miliseconds

Any explanations?

Thanks.

Dan
On 7/26/07, Daniel Marques <marques77 at gmail.com> wrote:
> Hi All,
>
> I'm running some quickie experiments with Spread and was wondering if
> anyone could help explain some performance results I'm seeing.
>
> Spread is configured as follows:
>
> Spread_Segment  192.168.143.255:4803 {
>        a 192.168.143.86
>        b 192.168.143.56
> }
>
> There are two Spread processes running, the 'server' on machine a, and
> the 'client' on machine b.  I'm running a ping-pong test: the client
> sends a message to the server, the server receives that message and
> responds back to the client.  They do this for 1024 messages, each 1KB
> in size.
>
> I'm seeing a huge disparity in the time of SAFE_MESS type messages and
> the others:
>
> UNRELIABLE_MESS: elapsed time is 2 second 562000 miliseconds
> RELIABLE_MESS: elapsed time is 2 second 547000 miliseconds
> FIFO_MESS: elapsed time is 2 second 547000 miliseconds
> CAUSAL_MESS: elapsed time is 2 second 531000 miliseconds
> AGREED_MESS: elapsed time is 2 second 547000 miliseconds
> SAFE_MESS : elapsed time is 274 second 141000 miliseconds
>
> Can anyone explain why it is so much slower?
>
> Thanks.
>
> Dan
>