On Fri, Aug 5, 2011 at 03:06, Yaxuan <[email protected]> wrote: > Although it is not difficult to implement Range-match on OvS, it is > definitly a bad thing for hardware-based OpenFlow switches.** > ** > ** >
Range-match should be an interface-provided convenience that your controller or northbound API should provide for you. Internally it would still gets converted to N individual entries. You could of course extend the protocol if you wanted to add range-match support in a single flow entry, but as you say, generally hardware doesn't support this match (although in theory things like NetFPGA and network-processor-driven data planes could). In the worst case, 1 range can expand to 2(W-1) prefixes, where W is the > bit-width of a header field (e.g. W=16 for L4 port fields). So for the TCAM > based hardware, it might waste many TCAM entries for a single rule. > It shouldn't waste any more than what you would actually do if you didn't have range functionality, however, and presumably the user knows what they're doing (har har, I guess). -- Nick
_______________________________________________ discuss mailing list [email protected] http://openvswitch.org/mailman/listinfo/discuss
