Comments on: Why interrupt affinity with multiple cores is not such a good thing http://www.alexonlinux.com/why-interrupt-affinity-with-multiple-cores-is-not-such-a-good-thing Sun, 05 Feb 2012 21:17:46 +0000 hourly 1 http://wordpress.org/?v= By: Alexander Sandler http://www.alexonlinux.com/why-interrupt-affinity-with-multiple-cores-is-not-such-a-good-thing/comment-page-1#comment-24654 Alexander Sandler Mon, 27 Dec 2010 08:33:06 +0000 http://www.alexonlinux.com/?p=1613#comment-24654 <a href='#comment-24645' rel="nofollow">@Mike Waychison</a> Good point. Thanks.<div class="comment-remix-meta"><a href="#" class="replyto" onclick="replyto('24654','Alexander Sandler'); return false;">Reply</a> - <a href="#" class="quote" onclick="quote('24654','Alexander Sandler','<a href=\'#comment-24645\' rel=\"nofollow\">@Mike Waychison<\/a>\r\nGood point. Thanks.'); return false;">Quote</a></div> @Mike Waychison
Good point. Thanks.
Reply – Quote
]]> By: Mike Waychison http://www.alexonlinux.com/why-interrupt-affinity-with-multiple-cores-is-not-such-a-good-thing/comment-page-1#comment-24645 Mike Waychison Sun, 26 Dec 2010 06:32:20 +0000 http://www.alexonlinux.com/?p=1613#comment-24645 RR interrupt delivery for received network packets can also cause congestion collapse of open streams. TCP assumes in its design that the underlying network will deliver packets _in order_. Spraying packets randomly across cores in a machine will look like intermittent failure on the network path and the only way TCP knows to handle the situation is to collapse the congestion window.<div class="comment-remix-meta"><a href="#" class="replyto" onclick="replyto('24645','Mike Waychison'); return false;">Reply</a> - <a href="#" class="quote" onclick="quote('24645','Mike Waychison','RR interrupt delivery for received network packets can also cause congestion collapse of open streams. TCP assumes in its design that the underlying network will deliver packets _in order_. Spraying packets randomly across cores in a machine will look like intermittent failure on the network path and the only way TCP knows to handle the situation is to collapse the congestion window.'); return false;">Quote</a></div> RR interrupt delivery for received network packets can also cause congestion collapse of open streams. TCP assumes in its design that the underlying network will deliver packets _in order_. Spraying packets randomly across cores in a machine will look like intermittent failure on the network path and the only way TCP knows to handle the situation is to collapse the congestion window.
Reply – Quote
]]> By: Alexander Sandler http://www.alexonlinux.com/why-interrupt-affinity-with-multiple-cores-is-not-such-a-good-thing/comment-page-1#comment-24497 Alexander Sandler Mon, 15 Nov 2010 17:07:19 +0000 http://www.alexonlinux.com/?p=1613#comment-24497 <a href='#comment-24447' rel="nofollow">@Bhushan</a> I think you should really reread the post. Also, take a look at related posts.<div class="comment-remix-meta"><a href="#" class="replyto" onclick="replyto('24497','Alexander Sandler'); return false;">Reply</a> - <a href="#" class="quote" onclick="quote('24497','Alexander Sandler','<a href=\'#comment-24447\' rel=\"nofollow\">@Bhushan<\/a>\r\nI think you should really reread the post. Also, take a look at related posts.'); return false;">Quote</a></div> @Bhushan
I think you should really reread the post. Also, take a look at related posts.
Reply – Quote
]]> By: Bhushan http://www.alexonlinux.com/why-interrupt-affinity-with-multiple-cores-is-not-such-a-good-thing/comment-page-1#comment-24447 Bhushan Tue, 02 Nov 2010 06:50:18 +0000 http://www.alexonlinux.com/?p=1613#comment-24447 Hi Alex, As you mentioned here throughput plays a major role here. So if in a system throughput is high along with some of the processes, then irq handling on some of the cores along with process affinity would help or not? I am also not sure what would be default process affinity of all the process, would all these processess served by cpu0.<div class="comment-remix-meta"><a href="#" class="replyto" onclick="replyto('24447','Bhushan'); return false;">Reply</a> - <a href="#" class="quote" onclick="quote('24447','Bhushan','Hi Alex,\r\n\r\nAs you mentioned here throughput plays a major role here. So if in a system throughput is high along with some of the processes, then irq handling on some of the cores along with process affinity would help or not? I am also not sure what would be default process affinity of all the process, would all these processess served by cpu0.'); return false;">Quote</a></div> Hi Alex,

As you mentioned here throughput plays a major role here. So if in a system throughput is high along with some of the processes, then irq handling on some of the cores along with process affinity would help or not? I am also not sure what would be default process affinity of all the process, would all these processess served by cpu0.

Reply – Quote
]]> By: Sharon Bruton http://www.alexonlinux.com/why-interrupt-affinity-with-multiple-cores-is-not-such-a-good-thing/comment-page-1#comment-23704 Sharon Bruton Thu, 28 Jan 2010 21:33:33 +0000 http://www.alexonlinux.com/?p=1613#comment-23704 Just keep posting good stuff.<div class="comment-remix-meta"><a href="#" class="replyto" onclick="replyto('23704','Sharon Bruton'); return false;">Reply</a> - <a href="#" class="quote" onclick="quote('23704','Sharon Bruton','Just keep posting good stuff.'); return false;">Quote</a></div> Just keep posting good stuff.
Reply – Quote
]]> By: ntop » IRQ Balancing http://www.alexonlinux.com/why-interrupt-affinity-with-multiple-cores-is-not-such-a-good-thing/comment-page-1#comment-23587 ntop » IRQ Balancing Sun, 13 Dec 2009 22:22:46 +0000 http://www.alexonlinux.com/?p=1613#comment-23587 [...] Why interrupt affinity with multiple cores is not such a good thing Author: admin Filed Under Category: PF_RING, TNAPI Article Comments: No Comments [...]<div class="comment-remix-meta"><a href="#" class="replyto" onclick="replyto('23587','ntop » IRQ Balancing'); return false;">Reply</a> - <a href="#" class="quote" onclick="quote('23587','ntop » IRQ Balancing','[...] Why interrupt affinity with multiple cores is not such a good thing Author: admin Filed Under Category: PF_RING, TNAPI Article Comments: No Comments [...]'); return false;">Quote</a></div> [...] Why interrupt affinity with multiple cores is not such a good thing Author: admin Filed Under Category: PF_RING, TNAPI Article Comments: No Comments [...]
Reply – Quote
]]> By: Alexander Sandler http://www.alexonlinux.com/why-interrupt-affinity-with-multiple-cores-is-not-such-a-good-thing/comment-page-1#comment-23553 Alexander Sandler Sun, 01 Nov 2009 08:01:38 +0000 http://www.alexonlinux.com/?p=1613#comment-23553 <a href='#comment-23547' rel="nofollow">@arkon</a> My point is simple. It is probably better to handle small number of packets per second on single core rather then on multiple cores. This is because handling packets on multiple cores will cause cache misses. So, indeed throughput plays a major role here.<div class="comment-remix-meta"><a href="#" class="replyto" onclick="replyto('23553','Alexander Sandler'); return false;">Reply</a> - <a href="#" class="quote" onclick="quote('23553','Alexander Sandler','<a href=\'#comment-23547\' rel=\"nofollow\">@arkon<\/a>\r\nMy point is simple. It is probably better to handle small number of packets per second on single core rather then on multiple cores. This is because handling packets on multiple cores will cause cache misses.\r\nSo, indeed throughput plays a major role here.'); return false;">Quote</a></div> @arkon
My point is simple. It is probably better to handle small number of packets per second on single core rather then on multiple cores. This is because handling packets on multiple cores will cause cache misses.
So, indeed throughput plays a major role here.
Reply – Quote
]]> By: arkon http://www.alexonlinux.com/why-interrupt-affinity-with-multiple-cores-is-not-such-a-good-thing/comment-page-1#comment-23547 arkon Thu, 29 Oct 2009 18:49:24 +0000 http://www.alexonlinux.com/?p=1613#comment-23547 <a href='#comment-23544' rel="nofollow">@Alexander Sandler</a> - But how can you say or justify that single core handling network is better than a few? In other words, you are saying that you prefer not being locked on the object, rather than handling concrete data? I don't see a solid proof, I think it should be measured by its throughput. Unless, I don't understand something clearly.<div class="comment-remix-meta"><a href="#" class="replyto" onclick="replyto('23547','arkon'); return false;">Reply</a> - <a href="#" class="quote" onclick="quote('23547','arkon','<a href=\'#comment-23544\' rel=\"nofollow\">@Alexander Sandler<\/a> - \r\nBut how can you say or justify that single core handling network is better than a few? In other words, you are saying that you prefer not being locked on the object, rather than handling concrete data?\r\n\r\nI don\'t see a solid proof, I think it should be measured by its throughput. Unless, I don\'t understand something clearly.'); return false;">Quote</a></div> @Alexander Sandler –
But how can you say or justify that single core handling network is better than a few? In other words, you are saying that you prefer not being locked on the object, rather than handling concrete data?

I don’t see a solid proof, I think it should be measured by its throughput. Unless, I don’t understand something clearly.

Reply – Quote
]]> By: Alexander Sandler http://www.alexonlinux.com/why-interrupt-affinity-with-multiple-cores-is-not-such-a-good-thing/comment-page-1#comment-23544 Alexander Sandler Thu, 29 Oct 2009 15:08:57 +0000 http://www.alexonlinux.com/?p=1613#comment-23544 <a href='#comment-23542' rel="nofollow">@arkon</a> Other cores won't be locked, waiting for some network object, because all work with network objects will be done on single core. This eliminates possibility of race.<div class="comment-remix-meta"><a href="#" class="replyto" onclick="replyto('23544','Alexander Sandler'); return false;">Reply</a> - <a href="#" class="quote" onclick="quote('23544','Alexander Sandler','<a href=\'#comment-23542\' rel=\"nofollow\">@arkon<\/a>\r\nOther cores won\'t be locked, waiting for some network object, because all work with network objects will be done on single core. This eliminates possibility of race.'); return false;">Quote</a></div> @arkon
Other cores won’t be locked, waiting for some network object, because all work with network objects will be done on single core. This eliminates possibility of race.
Reply – Quote
]]> By: arkon http://www.alexonlinux.com/why-interrupt-affinity-with-multiple-cores-is-not-such-a-good-thing/comment-page-1#comment-23542 arkon Tue, 27 Oct 2009 01:39:08 +0000 http://www.alexonlinux.com/?p=1613#comment-23542 Hey there, I was wondering, tell me if I'm wrong, that if you direct the interrupts to a single specific core from the network device (for instance), then still the other processors can be locked waiting for some network object, so I don't see how it's better. So either way other processors will end up waiting for the synch objects in order to access some data. However, it seems it might help cache coherency, but I am not sure.<div class="comment-remix-meta"><a href="#" class="replyto" onclick="replyto('23542','arkon'); return false;">Reply</a> - <a href="#" class="quote" onclick="quote('23542','arkon','Hey there,\r\n\r\nI was wondering, tell me if I\'m wrong, that if you direct the interrupts to a single specific core from the network device (for instance), then still the other processors can be locked waiting for some network object, so I don\'t see how it\'s better. So either way other processors will end up waiting for the synch objects in order to access some data.\r\n\r\nHowever, it seems it might help cache coherency, but I am not sure.'); return false;">Quote</a></div> Hey there,

I was wondering, tell me if I’m wrong, that if you direct the interrupts to a single specific core from the network device (for instance), then still the other processors can be locked waiting for some network object, so I don’t see how it’s better. So either way other processors will end up waiting for the synch objects in order to access some data.

However, it seems it might help cache coherency, but I am not sure.

Reply – Quote
]]>