[Spread-users] mod_log_spread and RELIABLE_MESS

Yair Amir yairamir at cnds.jhu.edu
Wed Apr 3 19:41:09 EST 2002


Hi,

It is not possible to create end-to-end (application-to-application) flow
control directly in the Spread toolkit without reverting to blocking the
whole system in the face of one slow receiver. We have discussed this issue
on the list before.

I agree that it might be useful to have *support* for end-to-end flow control
in the form of the library support that was discussed on this list also.

The question about unreliable versus reliable messages has nothing to
do with flow control actually. In general, there is not a big difference
between reliable and unreliable messaging in spread 3.x, especially
if it is run on a local area network. In a local area network, Spread's
internal flow control makes sure that messages are fairly rarely
missed, unless there is a problem in the hardware or a setup
problem. So, no - I am fairly sure there will be not much difference
in that case.

	Cheers,

	:) Yair.

John David Duncan wrote:

> 
> 
>> Daniel Rall wrote:
>>
>>> John David Duncan <jdd at greatschools.net> writes:
>>>
>>>> So here's a simple question:  is there any advantage in having
>>>> mod_log_spread send messages as UNRELIABLE_MESS rather than
>>>> RELIABLE_MESS?
>>>
>>> Indeed.  Without flow control built in to the Spread server, the
>>> "reliable" messages aren't nearly so reliable.
>>
>>
> 
> 
> Well, there is certainly some flow control built into spread, and there 
> might not be any good way for mod_log_spread (as a sender) to implement 
> application-level flow control, because of the way apache processes come 
> and go.
> 
> My problem isn't that my logging machine can't keep up -- it's that the 
> slowest webserver/sender occasionally gets bogged down, and that this 
> effects the rest of the segment.
> 
> What I'm asking, since I don't need the semantics of reliable messages, 
> is whether it actually makes life any easier for spread if I send 
> unreliable ones.
> 
> 
> 
>> I have mod_log_spread logging fine in a decent size production cluster 
>> (4 million hits/day) and see almost no message loss on the Spread 
>> level, so even if I did change the RELIABLE_MESS to UNRELIABLE_MESS, I 
>> would have only lost a handful (<100) messages over the past week.
>>
>> If you are seeing message loss on the link layer or are overflowing 
>> your recv buffers, then you need to investigate the way you have 
>> Spread set up.
> 
> 
> 
> 
> _______________________________________________
> Spread-users mailing list
> Spread-users at lists.spread.org
> http://lists.spread.org/mailman/listinfo/spread-users
> 
> 







More information about the Spread-users mailing list