[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openvswitch-discuss
Subject:    [ovs-discuss] Is flow extraction on the critical path
From:       jpettit () nicira ! com (Justin Pettit)
Date:       2012-01-20 8:39:07
Message-ID: 07250821-B3B6-480A-A79A-1FCA7940FEEF () nicira ! com
[Download RAW message or body]

It depends on the port, but any reasonable implementation would do the parsing in \
hardware.  Current merchant switching ASICs can't typically match on the full set of \
fields defined in OpenFlow, so often times a hardware port will only offer line rate \
for a subset.

--Justin


On Jan 19, 2012, at 9:25 PM, K.Kalapriya wrote:

> hi, 
> 
> One more query, Is this flow extraction typically done by a software module even in \
> hardware switches or it is perhaps performed by a specialized and optimized \
> hardware in hardware switches? 
> regards,
> kalapriya.
> 
> From: Justin Pettit <jpettit at nicira.com>
> To: K.Kalapriya <kalapriya at yahoo.com> 
> Cc: "discuss at openvswitch.org" <discuss at openvswitch.org> 
> Sent: Thursday, January 19, 2012 3:05 PM
> Subject: Re: [ovs-discuss] Is flow extraction on the critical path
> 
> Open vSwitch doesn't support OpenFlow 1.1, but, regardless, flow extraction is on \
> the critical path.  I can't think how to avoid that, since it must classify the \
> packets to determine which actions to apply.  It does some smart caching to make \
> the lookup fast in software, but it still needs to extract the flow to do that \
> lookup. 
> --Justin
> 
> 
> On Jan 19, 2012, at 1:27 AM, K.Kalapriya wrote:
> 
> > hi,
> > 
> > Just wanted to know if the flow extraction phase (page 9 of the openflow switch \
> > specification 1.1.0) on the critical path of the switch performance? 
> > regards,
> > kalapriya.
> > _______________________________________________
> > discuss mailing list
> > discuss at openvswitch.org
> > http://openvswitch.org/mailman/listinfo/discuss
> 
> 
> 


[prev in list] [next in list] [prev in thread] [next in thread]