<div dir="ltr"><div>Thanks for the kind reply !!</div><div><br></div><div>Have a nice day !!</div></div><div class="gmail_extra"><br><div class="gmail_quote">2017-06-09 14:37 GMT+09:00 Ben Pfaff <span dir="ltr"><<a href="mailto:blp@ovn.org" target="_blank">blp@ovn.org</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">With microflow cache only, there's one cache miss per microflow.<br>
With an effective megaflow cache, the miss rate is much lower.<br>
<br>
The paper gives examples for your other question.<br>
<div class="HOEnZb"><div class="h5"><br>
On Fri, Jun 09, 2017 at 01:47:14PM +0900, Heung Sik Choi wrote:<br>
> Thanks to reply.<br>
><br>
> Regarding your second question, which flow rules are you asking about?<br>
><br>
> I just want to know that when ovs had only an in-kernel microflow cache,<br>
> How many in-kernel cache miss there are, and also how much improving<br>
> there are when using megaflow.<br>
><br>
> And I have a last question. short-lived flows were main problem when there<br>
> was micro flow cache. I don't know the circumstance where short lived<br>
> connection happen often. can you let me know about it?<br>
><br>
><br>
> Thanks!!<br>
><br>
><br>
> 2017-06-09 13:29 GMT+09:00 Ben Pfaff <<a href="mailto:blp@ovn.org">blp@ovn.org</a>>:<br>
><br>
> > When OVS had only an in-kernel microflow cache, there were at least two<br>
> > reasons for performance problems with many short-lived flows. The first<br>
> > was the cost of sending packets to userspace. The second was the cost<br>
> > of translating the packets through the entire OpenFlow pipeline. The<br>
> > megaflow cache solves both problems: it eliminates both the<br>
> > kernel-to-user-to-kernel transition and (if the megaflow cache is<br>
> > effective) the additional OpenFlow translations.<br>
> ><br>
> > Regarding your second question, which flow rules are you asking about?<br>
> ><br>
> > On Fri, Jun 09, 2017 at 12:22:09PM +0900, Heung Sik Choi wrote:<br>
> > > What does "if only the microflow cache works" mean?<br>
> > ><br>
> > > sorry to confuse you. I'm not good at English.<br>
> > ><br>
> > > In the paper, the authors say that at start of ovs implementation, there<br>
> > > has been microflow cache(EMC), but megaflow wasn't implemented.<br>
> > ><br>
> > > At that time, they say there was a problem caused by short lived<br>
> > > connections.<br>
> > ><br>
> > > I want to know the problem detail. I guess that the problem made ovs get<br>
> > > flow rules from userlevel, and that made performance degradation.<br>
> > ><br>
> > > also I want to know that when not using megaflow, are there big<br>
> > difference<br>
> > > in aspect of number of flow rules.If there are the big difference, can<br>
> > you<br>
> > > tell me the difference in numbers?<br>
> > ><br>
> > ><br>
> > ><br>
> > > 2017-06-09 0:43 GMT+09:00 Ben Pfaff <<a href="mailto:blp@ovn.org">blp@ovn.org</a>>:<br>
> > ><br>
> > > > On Thu, Jun 08, 2017 at 04:33:54PM +0900, Heung Sik Choi wrote:<br>
> > > > > 1. If only microflow cache works and there are many short lived<br>
> > > > > connections, does it make many tuples in table, and does it suffers<br>
> > > > serious<br>
> > > > > performance degradation by the many tuples(very many tuple makes<br>
> > context<br>
> > > > > switching to Userlevel)?<br>
> > > ><br>
> > > > What does "if only the microflow cache works" mean?<br>
> > > ><br>
> > > > > 2. I know that OVS uses hash for table index. if only microflow<br>
> > cache and<br>
> > > > > there are small amount of entries in the table, is it possible to<br>
> > find<br>
> > > > > entry in O(1) time?<br>
> > > ><br>
> > > > Yes.<br>
> > > ><br>
> > > > > 3. In paper, microflow and megaflow use the word 'cache'. However,<br>
> > when I<br>
> > > > > open the OVS code, it looks like they are in main memory area.<br>
> > doesn't<br>
> > > > the<br>
> > > > > 'cache' mean hardware cache?<br>
> > > ><br>
> > > > No.<br>
> > > ><br>
> ><br>
</div></div></blockquote></div><br></div>