Howdy!
We're developing an application based on the 82599 and I'm hoping that an
expert can check our understanding.
Our applications runs with multiple independent processes for multi-core
scalability. We want to distribute the traffic arriving on the 82599
between this processes in a nice way. For us "nice" means that all traffic
to and from a given client's IP address will be dispatched to the same
process. ("Clients" being a set of machines in a particular subnet -- let's
suppose it's a /20.)
So packets with a source address in the client subnet should be
hash-dispatched on source address to an RX queue, and packets with a
destination address in the client subnet should be identically
hash-dispatched on the destination address.
Our understanding based on study of the 82599 data sheet is that neither
RSS nor Flow Director is able to accomplish this, and so we have to do our
dispatching in software instead of hardware. Can someone please confirm
whether this is correct? (If so is there anything on Intel's roadmap that
could help us out?)
Thanks very much!
-Luke
------------------------------------------------------------------------------
Keep Your Developer Skills Current with LearnDevNow!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-d2d
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel® Ethernet, visit
http://communities.intel.com/community/wired
