Maximum message size 2006-04-04 12:18:37 <Chad>

[Branden C. Christensen]

Maximum message size 2006-04-04 12:18:37 <Chad>




Hi all,

I wrote an Earthworm import module (slink2ew) that creates TRACEBUF[2]
messages up to 2984 bytes in length. Since this is well below the
maximum of
4096 bytes I thought all was safe. But every now and then I hear from
slink2ew users that have small maximum message sizes set for
wave_serverV
(e.g. the default 1064 MaxMsgSize in wave_serverV.d) and the slink2ew
generated messages are rejected.

What are the consequences of changing the maximum message sizes for
all ring
clients to 4096? Does this, for example, require wave_serverV to use
more
disk space?

thanks,
Chad
IRIS-DMC






Re: Maximum message size 2006-04-04 13:17:55 <mwithers@...>




If I understand correctly. The MaxMsgSize just governs the size of
of the incoming message buffer. It should be as large as the largest
tracebuf waveserver will see on the memory ring its attached to. These
messages don't necessarily get written to the tank file on disk. This
parameter should not be confused with the channel specific Record size
parameter. Any give message written to the tank file can't be larger
than Record size configured for that tank. And all trace snippets will
take one record size chunk of disk regardless of whether or not
they need the entire chunk.

On Tue, 4 Apr 2006, Chad wrote:

>
> Hi all,
>
> I wrote an Earthworm import module (slink2ew) that creates TRACEBUF[2]
> messages up to 2984 bytes in length. Since this is well below the maximum of
> 4096 bytes I thought all was safe. But every now and then I hear from
> slink2ew users that have small maximum message sizes set for wave_serverV
> (e.g. the default 1064 MaxMsgSize in wave_serverV.d) and the slink2ew
> generated messages are rejected.
>
> What are the consequences of changing the maximum message sizes for all ring
> clients to 4096? Does this, for example, require wave_serverV to use more
> disk space?
>
> thanks,
> Chad
> IRIS-DMC
>
>

Mitch

Center for Earthquake Research and Information (CERI)
University of Memphis Ph: 901-678-4940
Memphis, TN 38152 Fax: 901-678-4734







Re: Maximum message size 2006-04-04 15:41:54 <David Kragness>



Chad,
The primary consequence of changing MaxMsgSize from 1064 to 4096 is a
bigger memory footprint
for the wave server, but the change should be fairly neglegable
unless
you are strapped for RAM.
The primary memory consumption relative to MaxMsgSize is the
input queue that's used to hold tracebuf messages coming off the
ring.
The size of memory used by the queue is MaxMsgSize * InputQueueLen.
So even if you were running with an outrageously large
InputQueueLen=2000, then changing the
MaxMsgSize from 1064 to 4096 would only cause a memory footprint
increase of
(4096-1064) * 2000 = 6MB.

good luck,
david

Chad wrote:

>Hi all,
>
>I wrote an Earthworm import module (slink2ew) that creates TRACEBUF[2]
>messages up to 2984 bytes in length. Since this is well below the maximum of
>4096 bytes I thought all was safe. But every now and then I hear from
>slink2ew users that have small maximum message sizes set for wave_serverV
>(e.g. the default 1064 MaxMsgSize in wave_serverV.d) and the slink2ew
>generated messages are rejected.
>
>What are the consequences of changing the maximum message sizes for all ring
>clients to 4096? Does this, for example, require wave_serverV to use more
>disk space?
>
>thanks,
>Chad
>IRIS-DMC