I would go with an FPGA approach, if I wanted to implement a full OF compliant switch. Software only switches are easier to implement yes, but very limited in terms of performance. I would use a software approach only If I wanted to implement an 100Mb switch of maybe, a 1Gb switch with low port number.
On 27 March 2014 02:15, Michael Orr <o...@marvell.com> wrote: > I'll jump in- > Switches need to have firmware that supports OpenFlow to make it usable. > Either the vendor adds it in the firmware you get when you buy it, or you > can upgrade the firmware - if the vendor has an updated one that supports > OF. > AS for wire-speed yes or no : it varies with the switch, and what you are > trying to make it do. > Switches that use "merchant silicon" will have limitations in both the > number of rules supported and the nature of the rules supported (usually, > matching is no problem, but supporting multi-stage very flexible actions > can be problematic) If you aks the switch to do too many rules, or even one > rule that is too complex to handle in the hardware, the vendor has two > choices: Refuse the rule and tell the controller "not supported" or > implement it in software, which *for these rules only* will get you > less-than-wire-speed. > What you can do: > 1. You may not NEED wire speed for your real-life applications. Out of the > lab, very few applications really generate back-to-back wire speed packets > 2. Use a very sophisticated (and usually more expensive) NPU based switch > (E.g. From Netronome) that will give you full OpenFlow at wire speed. > > (There are, of course, SW-Only switches, that do not have any real > switching HW - these are essentially linux computes that receive packets > through their NW adapter(s) process it in software on the CPU, and send > the result back out through a NW adapter) > > -- > Michael Orr. Marvell CSI BU - System Architecture Cell:(408)636-3744 > Note: Dyslexic here. Even on speller-equipped iPad, typos Probable. > > On Mar 25, 2014, at 7:37 PM, "durga" <c.vijaya.du...@gmail.com> wrote: > > Sounds legit, I know this is a pretty old (outdated) article, but > highlights the consequences of OF on software and OF on hardware > > > http://searchsdn.techtarget.com/news/2240150716/Brocade-Some-OpenFlow-switches-cant-operate-at-line-rate-speeds > > May be someone from the core group throw some light on how the existing > legacy devices can be made compatible with Openflow? does it really need a > firmware upgrade or a software upgrade is sufficient? > > Cheers! > Durga > > > > On Wed, Mar 26, 2014 at 1:18 PM, Naman Muley <naman.g.mu...@gmail.com>wrote: > >> I'd reckon that is not completely true. >> >> A lot of switches, just to reach to their customers quickly, implement >> openflow in software. That means, openflow actions and everything are >> implemented in software and no specialized ASICs are added to implement the >> functionality. This of course obstructs line rate and the performance is >> observably slower than the ones that implement openflow in hardware. >> >> An example is, you have a flow entry that rewrites IP addresses for every >> packet that matches the criteria. This, if done with hardware will give you >> line rate performance. But if done in software, will make it noticeably >> slower. >> >> Please correct if I'm wrong. >> >> Naman >> On Mar 25, 2014 10:10 PM, "durga" <c.vijaya.du...@gmail.com> wrote: >> >>> in continuation of the topic, >>> I had remotely read, that enabling OpenFlow is as much as as doing a >>> firmware upgrade, can someone confirm it? or the only way to have OpenFlow >>> environment is by replacing legacy devices with brand-new openflow enabled >>> devices? >>> >>> Cheers! >>> Durga >>> >>> >>> >>> On Wed, Mar 26, 2014 at 12:18 AM, Carlos Ferreira <carlosmf...@gmail.com >>> > wrote: >>> >>>> Then if it's not Openflow capable, then you cannot use Openflow with it. >>>> >>>> >>>> On 24 March 2014 19:31, Andrew Niteesh <andrewec...@gmail.com> wrote: >>>> >>>>> No the equipment is not an OpenFlow capable switch. >>>>> >>>>> >>>>> On Mon, Mar 24, 2014 at 12:46 PM, Carlos Ferreira < >>>>> carlosmf...@gmail.com> wrote: >>>>> >>>>>> What do you mean Enable? Please clarify. Is the equipment, an >>>>>> OpenFlow capable Switch? >>>>>> >>>>>> >>>>>> On 17 March 2014 19:31, Andrew Niteesh <andrewec...@gmail.com> wrote: >>>>>> >>>>>>> Hi, >>>>>>> >>>>>>> I have a small clarification, could you please tell me how to enable >>>>>>> openflow in Ethernet Switch? >>>>>>> >>>>>>> Regards, >>>>>>> Andrew Niteesh >>>>>>> 8324179269 >>>>>>> >>>>>>> _______________________________________________ >>>>>>> openflow-discuss mailing list >>>>>>> openflow-discuss@lists.stanford.edu >>>>>>> https://mailman.stanford.edu/mailman/listinfo/openflow-discuss >>>>>>> >>>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>> >>>>>> Carlos Miguel Ferreira >>>>>> Researcher at Telecommunications Institute >>>>>> Aveiro - Portugal >>>>>> Work E-mail - c...@av.it.pt >>>>>> Skype & GTalk -> carlosmf...@gmail.com >>>>>> LinkedIn -> http://www.linkedin.com/in/carlosmferreira >>>>>> >>>>> >>>>> >>>> >>>> >>>> -- >>>> >>>> Carlos Miguel Ferreira >>>> Researcher at Telecommunications Institute >>>> Aveiro - Portugal >>>> Work E-mail - c...@av.it.pt >>>> Skype & GTalk -> carlosmf...@gmail.com >>>> LinkedIn -> http://www.linkedin.com/in/carlosmferreira >>>> >>>> _______________________________________________ >>>> openflow-discuss mailing list >>>> openflow-discuss@lists.stanford.edu >>>> https://mailman.stanford.edu/mailman/listinfo/openflow-discuss >>>> >>>> >>> >>> _______________________________________________ >>> openflow-discuss mailing list >>> openflow-discuss@lists.stanford.edu >>> https://mailman.stanford.edu/mailman/listinfo/openflow-discuss >>> >>> > _______________________________________________ > openflow-discuss mailing list > openflow-discuss@lists.stanford.edu > https://mailman.stanford.edu/mailman/listinfo/openflow-discuss > > > _______________________________________________ > openflow-discuss mailing list > openflow-discuss@lists.stanford.edu > https://mailman.stanford.edu/mailman/listinfo/openflow-discuss > > -- Carlos Miguel Ferreira Researcher at Telecommunications Institute Aveiro - Portugal Work E-mail - c...@av.it.pt Skype & GTalk -> carlosmf...@gmail.com LinkedIn -> http://www.linkedin.com/in/carlosmferreira
_______________________________________________ openflow-discuss mailing list openflow-discuss@lists.stanford.edu https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
