[OLSR-users] Plugin, several net_output calls at once.
Mon Aug 16 13:42:06 CEST 2004
Hi again Andreas.
I am having an interesting observation. I am sending packets in my plugin with lots of information.
If this information should happen to exceed the max size of 512, I call a net_output, rebuild the header and add the remaining
data before calling net_output again and so on until the last message does not exceed the max.
I take care to completely rebuild the buffer and olsr message header, with new sequence numbers and everything.
To test this I have just defined a max size of 115 for my testing purposes in my plugin, NOT olsr!
This will allow my data to be sent in 2 messages, the first message containing 2 units of data, and the last message containing one unit of data.
My solution works well on 2 machines, but the 3rd machine, which is quite slow, will buffer all three units of data and send them with a size
of something like 148, which is over my artificial max size of 115.
My question is therefore, does the core OLSR buffer the net_output messages on slow machines, or congested lines?
Especially if the size does not exceed the 512 limit of several net_output calls?
Is there a place in the OLSR code that I can restrict the max message size to 115 to test if it is OLSR that buffers the messages?
I know my test is a bit porked since OLSR still operates with a max size of 512, while I have artificially restricted my plugin to 115.
How can I best test this, and am I right at all about my net_output buffering assumption, or do you think the problem is something else?
If you think this should work if the size is restricted in OLSR and not the plugin, where is the size restriction in OLSR?
Thales Communications Norway
More information about the Olsr-users