On Tue, Jan 29, 2013 at 2:28 PM, Steve Noble <sno...@sonn.com> wrote: > This is a tough situation. Very few of us are happy with how > different vendors are putting different types of traffic into SW vs HW > flows. I hold a belief that all flows should go into HW until you run > out of HW space, then flows should go into SW. At least in that > scenario you can limit the # of flows you enter into the box and not > hit performance issues.
It is unlikely that a vendor would elect to put a flow into software (other than a few switches which have user-configurable knobs) when the switch is capable of processing the flow in hardware. The issue here is what is capable of being *programmed* to be matched in hardware. Obviously the 3500 is capable of L2 forwarding in hardware, but the table may not be programmable in a way that is consistent with the OpenFlow API. > Pica8 handles > 4k flows (I tested up to 4700 or so) on their lower > end (329x) switches and I believe up to 16k in their upper end > switches (3920) (I can't find a link). This is a bit of an oversimplification - as with all switches it depends on what *type* of flows. You are likely to find that the switch supports significantly fewer L3 match flowmods than L2 match flowmods. > The difference between Pica8 and some other vendors is that Pica8 has > a OVS mode where you don't boot their normal PicaOS and just use OVS > to run the switch. This allows for a lot more flows. In Hybrid > (PicaOS) mode, the numbers are lower, around 1700 if I remember > correctly. This is the right number to compare with fully hybrid > switches such as the HPs. The real difference in the case of a 329x versus an HP 3500 is the switch chip, not the software. The switch is not capable of various matches and actions in hardware simply because it is running OVS - it is because the switch chip itself supports these matches and actions. Yes, software can limit a device to less than the ASIC is capable of (or shave off some resources for non-OF features, as PicaOS mode does), but it can't expand what can be done in hardware beyond the capability of the chip. In this case the question of flows being punted to software in the 3500 has nothing to do with table size (as I suspect his table isn't anywhere near full), and everything to do with what flow matches (and actions) can be programmed into the flow table at all. -- Nick _______________________________________________ openflow-discuss mailing list openflow-discuss@lists.stanford.edu https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
