|Re: [CDO-Net4j] Measuring bandwidth of Net4j+CDO [message #986229 is a reply to message #986226]
||Mon, 19 November 2012 14:24
| Guido Scalise
Registered: April 2012
I thought about checking it with Wireshark, too... after all, what I'd |
like to know is the average b/w needed for our application's operations,
so I can have a better understanding on our app's b/w requirements and
discover any limitations in infrastructure due to these requirements.
OTOH, having this at a Net4J level would help to better isolate and
identify different scenarios that could appear... for instance, it might
be interesting to identify b/w due to delta notifications vs b/w due to
model access/modification, and then make some tests to know how this
might be affected by the number of concurrent clients working on a model.
I could create the dissector, it doesn't seem hard at all, but I don't
know if the info I could obtain from it would be as useful as having the
same info directly on net4j.
El 19/11/2012 15:10, Christophe Bouhier escribió:
> I recently looked into this a bit (To get an understanding of B/W as
> well). I tried a couple of things. First you can turn on tracing on
> Net4j, this will dump information on channels, streams, buffers. So it
> could give you an ID of what is happening. You would need to backward
> engineer the dumps, to make statistics out of it.
> I also took a trace with Wireshark on the port 2036 (Which is the
> default I believe for TCP). Wireshark coudn't interpret the Net4J, but
> perhaps it would be doable to create such a Wireshark dissector:
> (It's C, so a bit of IDE setup would be needed).
> What would be needed is some basic description of is to be expected on
> the wire. Cheers Christophe
Powered by FUDForum
. Page generated in 0.03194 seconds