BIND 10 #245: Meaure performance of various "receiptionist" designs

[BIND 10 Development]

#245: Meaure performance of various "receiptionist" designs
----------------------+-----------------------------------------------------
 Reporter:  shane     |       Owner:  stephen                    
     Type:  task      |      Status:  new                        
 Priority:  major     |   Milestone:  06. 4th Incremental Release
Component:  recurser  |    Keywords:                             
Sensitive:  0         |  
----------------------+-----------------------------------------------------
 We need to see what the performance implications of various possible
 designs for receiving queries and directing them may be.

 Notes from the BIND 10 Face to Face meeting:

 {{{
 Shane - I guess the first thing might be to build some dummy programs for
 the receptionist.

 We need to benchmark various scenarios.

 how long to wake up

 how long to process many queries without sleeping between ultimately do we
 care how long it takes to wake up if it doesn't affect throughput?

 if waking up is more overhead than doing work, it could be a problem

 Shane wants: listening on a UDP socket, receiving a 50k packet, and doing
 a checksum.

 Also: take the data, pass it to another process and do it again.

 Also: queue up 1 work item, send, queue 2, send, queue 4, send 16, etc etc
 65536
 }}}

 Diagram from Michael:

 http://www.flame.org/~explorer/bind10-receptionist-test.png

-- 
Ticket URL: <http://bind10.isc.org/ticket/245>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development