Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

> >>-Original Message- >>> >>> [Wang, Yipeng] In my test, I compared the proposed EMC with current >EMC with same 16k entries. >>> If I turned off THP, the current EMC will cause many TLB misses because of >its larger entry size, which I profiled with vTunes. >>> Once I turned on THP with

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

>-Original Message- >[Wang, Yipeng] >Yes, with 2-hash function and 16 entries per bucket, CD has much less misses >in general. > >As first step to combine both CD and DFC, I incorporated the signature and >way-associative structure from CD into DFC. I just did >simple prototype without

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

Hi, Jan, Please find my reply inlined >-Original Message- >> >> [Wang, Yipeng] In my test, I compared the proposed EMC with current EMC with >> same 16k entries. >> If I turned off THP, the current EMC will cause many TLB misses because of >> its larger entry size, which I profiled with

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

>-Original Message- >> I think this is an interesting idea. >> >> One caveat is that in this case we use the rules' field to infer the flows' >> field. If the rule >> does not consider a field within which the flow has high entropy, it still >> does not help. >> Or if the rule considers

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

Hi Yipeng, On Mon, Jan 8, 2018 at 1:54 PM, Wang, Yipeng1 wrote: > I think this is an interesting idea. > > One caveat is that in this case we use the rules' field to infer the flows' > field. If the rule > does not consider a field within which the flow has high entropy,

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

I think this is an interesting idea. One caveat is that in this case we use the rules' field to infer the flows' field. If the rule does not consider a field within which the flow has high entropy, it still does not help. Or if the rule considers many fields but the flow does not have high

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

On Thu, Jan 4, 2018 at 5:41 PM, Wang, Yipeng1 wrote: > I agree with you guys on the issue with RSS hash. > > RSS is not originally designed for hash table so it is not flexible enough to > be configured > in the NICs. Smart NICs will be more capable to offload hash

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

Hi Yipeng, Thank you very much for your comments. Please find my answers inline. Regards, Jan > -Original Message- > From: Wang, Yipeng1 [mailto:yipeng1.w...@intel.com] > Sent: Wednesday, 20 December, 2017 00:09 > To: Jan Scheurich ; d...@openvswitch.org >

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

t;charlie....@intel.com> >Subject: RE: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache > >Thanks William for your comments. >Some replies inline. > >-Antonio > >> -Original Message- >> From: ovs-dev-boun...@openvswitch.org [mailto:ovs-dev- &

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

intel.com> > Cc: d...@openvswitch.org; Tai, Charlie <charlie@intel.com> > Subject: Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache > > Hi Yipeng, > > Thanks for the reply. > > On Tue, Jan 2, 2018 at 10:45 AM, Wang, Yipeng1 <yipeng1.w.

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

Hi Yipeng, Thanks for the reply. On Tue, Jan 2, 2018 at 10:45 AM, Wang, Yipeng1 wrote: > Hi, William, > > Thanks for the comment. You are right. If the RSS hash does not consider the > fields that matter, the situation you mentioned could happen. > > There are two

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

Hi, William, Thanks for the comment. You are right. If the RSS hash does not consider the fields that matter, the situation you mentioned could happen. There are two design options for CD as you may find, when the signatures collide, we could either replace the existing entry (the current

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

On Tue, Dec 19, 2017 at 3:08 PM, Wang, Yipeng1 wrote: > Hi, Jan, > > Please see my comments inlined. > > Thanks > Yipeng > >>-Original Message- >>From: Jan Scheurich [mailto:jan.scheur...@ericsson.com] >>Sent: Monday, December 18, 2017 9:01 AM >>To: Wang, Yipeng1

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

Hi Yipeng, Thanks a lot for your feedback. Please find some responses below. Regards, Jan From: Wang, Yipeng1 [mailto:yipeng1.w...@intel.com] Sent: Sunday, 17 December, 2017 19:49 To: Jan Scheurich ; d...@openvswitch.org Cc: Gobriel, Sameh ;

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

s-dev-boun...@openvswitch.org [mailto:ovs-dev- > > boun...@openvswitch.org] On Behalf Of Jan Scheurich > > Sent: 20 November 2017 17:33 > > To: d...@openvswitch.org > > Subject: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache > > > > So far t

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

ubject: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache > > So far the netdev datapath uses an 8K EMC to speed up the lookup of > frequently used flows by comparing the parsed packet headers against the > miniflow of a cached flow, using 13 bits of the packet RSS hash as in

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

Hi Billy, Thanks for your feedback. I continue to have problems with sending out patches via mail from our company network. This time it seems it is the "Page Feed" (^L) character in the dpif-netdev.c file that got corrupted in the emailed version of the patch. I wish there was a way to

Re: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

> boun...@openvswitch.org] On Behalf Of Jan Scheurich > Sent: Monday, November 20, 2017 5:33 PM > To: d...@openvswitch.org > Subject: [ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache > > So far the netdev datapath uses an 8K EMC to speed up the lookup of frequently &

[ovs-dev] [PATCH] dpif-netdev: Refactor datapath flow cache

So far the netdev datapath uses an 8K EMC to speed up the lookup of frequently used flows by comparing the parsed packet headers against the miniflow of a cached flow, using 13 bits of the packet RSS hash as index. The EMC is too small for many applications with 100K or more parallel packet flows