Bug 639803

Thanks, Don the content looks good, 1 request though: Could we please change the heading of the section to something other than "The Multicast Problem".  I know I described it as such in my writing to you, but it reads like we're shipping known defects when I see it in print for customer consumables.  Perhaps something more neutral, like "multicast performance", or "Multicast considerations"?

In answer to your questions, yo udistinguish which technique mitigates multicast performance the same way you do any other performance problem: By first looking at the metrics you've outlined in other sections and observing where frames are dropping.  The kernel doesn't care that its muticast traffic, its just dropping frames, and based on the location, you can take action to mitigate that.

In regards to your last sentence, I think its good, but It would be more correct to say:
...Alternatively, you can try optimizing applications use of a socket.."

Its the application that uses a socket, not a socket that uses an application.

(In reply to comment #7)
> Thanks, Don the content looks good, 1 request though: Could we please change
> the heading of the section to something other than "The Multicast Problem".  I
> know I described it as such in my writing to you, but it reads like we're
> shipping known defects when I see it in print for customer consumables. 
> Perhaps something more neutral, like "multicast performance", or "Multicast
> considerations"?
> 

you're right. Multicast Considerations sounds much better. revised as such.h

> In answer to your questions, yo udistinguish which technique mitigates
> multicast performance the same way you do any other performance problem: By
> first looking at the metrics you've outlined in other sections and observing
> where frames are dropping.  The kernel doesn't care that its muticast traffic,
> its just dropping frames, and based on the location, you can take action to
> mitigate that.
> 

ok

> In regards to your last sentence, I think its good, but It would be more
> correct to say:
> ...Alternatively, you can try optimizing applications use of a socket.."
> 
> Its the application that uses a socket, not a socket that uses an application.

thanks, revised as:
<new>
Alternatively, you can try optimizing an application's socket use; to do so, configure the application to control a single socket and disseminate the received network data quickly to other user-space processes.
</new>

setting this bug now to MODIFIED. thanks again, Neil!

Verified in Red_Hat_Enterprise_Linux-Performance_Tuning_Guide-6-en-US-1.0-28