On 03.10.2016 11:45, kp kirchdoerfer wrote: > Am Montag, 3. Oktober 2016, 10:41:05 schrieb Andrew: >> On 02.10.2016 23:37, Erich Titl wrote: >>> Hi Andrew >>> >>> Am 02.10.2016 um 19:58 schrieb Andrew: >>>> On 02.10.2016 20:45, Erich Titl wrote >>>> >>> :... >>> : >>>>> We need to find out if this is an issue for the majority of our users. >>>>> most of the drivers will probably just work as they work in standad >>>>> distros. >>>> This is issue for everybody who uses intel server cards in high-load >>>> applications. >>> Yes and this is why I asked if this was an issue for the _majority_ >> I don't know statistics how much people uses LEAF as home box/small >> office router, and how much people uses LEAF somewhere in production >> environment with high traffic. >> I know at least some people who uses LEAF in production (access >> servers/borders/etc), and I know only one case of usage as office >> router(just because soho router in that place frequently hangs). >> >>> ... >>> >>>> Actually intel NIC card is the only available choice for high-troughput >>>> routing. Other cards have too high CPU usage, or starts to drop packets >>>> at 60-70% of bandwidth usage. >>> So on a Gbit card you would have a throughput of roughly 600 Mbit. For >>> most users and I said _most_ users they will never be able to buy that >>> much power. >>> >>> And unless you have a massive parallel system you will have difficulties >>> to pass that much data through any kind of traffic management. I >>> observed issues in that aera, never at NIC level. >>> >>> So - yes, if it isn't a home routing box, >>> >>>> it uses Intel NIC. >>> Still the same question, is this an issue for the majority and therefore >>> a killer criterion? It might well be so, but I would like to know numbers. >>> >>> I for once do not have any intel NICs in usage, but yes, I did use them >>> a few years back and might have been happy to have hight troughput, but; >>> even then I never had a saturation issue at hand. YMMV >>> >>> Still I would not consider it a killer criterion, but yes, if someone is >>> willing to put in the effort to overcome limitations in that area, >>> great. I understand that you are working in a high speed/throughput >>> environment and there I see of course the usefulness of having >>> specialized drivers, but I _guess_ the majority of our users are not >>> limited there. They have problems in the integration of some of our >>> packages as seen lately in the leaf-user list. >>> >>> cheers and yes, please if you have spare time to integrate those >>> drivers, go for it. >>> >>> ET >> It isn't too hard to integrate them. I'll try to update drivers to >> latest version in master (which is 4.4-based), because there's no >> 4.7-based branch in git. > This was meant as testing what has to be done when moving to a newer kernl, > prefrrably a LTS as well. > And it occured that migrations issues will be with e1000e and igb. > > There will be enough time to solve once we move on. > > kp Hi.
Drivers update seems to be trivial - I just placed new archives, and removed some compat patches (that were added because driver fails to built with 4.4 kernel). If all will be built OK, I'll push changes. ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot _______________________________________________ leaf-devel mailing list leaf-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/leaf-devel
