>I don't really have the skill set That's terrifying. What kind of mad scientist would you need to be to work with it?!
Josh Luthman Office: 937-552-2340 Direct: 937-552-2343 1100 Wayne St Suite 1337 Troy, OH 45373 On Wed, Apr 22, 2015 at 1:20 PM, Simon Westlake <[email protected]> wrote: > That's about it, but two companies in that space selling complete Linux > networking stacks are 6WIND and Wind River, I've investigated them both > pretty extensively. It'd be fun to play with one day. Go look into the > Intel DPDK. The problem is you have to build each application to support > it, so I really doubt there's going to be much useful open source > development in this space, as everything has to be tailored to exactly what > you want it to do. And it is really expensive. I am pretty sure I had to > sign an NDA with both 6WIND and Wind River, so I won't throw specific > numbers out, but it's definitely not pocket change. > > I really wanted to do some stuff with it for the Powercode BMU, but I > don't really have the skill set (or the time right now) to work on it. > You're right though, it's very interesting. > > On 04/22/2015 12:06 PM, Josh Reynolds wrote: > > Sandvine looks to have similar specs on their platforms as well. > > Anyways, that would be really surprising to me Simon. I didn't expect a > multi-generational leap in performance until more things used PF_RING from > ntop, or things like netmap gain in development and popularity. I know > 6WIND does similar things with kernel bypass, pushing the stack into user > space, but AFAIK there are only about 4 or 5 companies with any sort of > kernel bypass capability of the network stack. > > If you have any additional information, please do share. This is a > fascinating topic I've been monitoring since around 2011. > > On April 22, 2015 8:54:37 AM AKDT, Simon Westlake <[email protected]> > <[email protected]> wrote: >> >> 600Gbps in software is actually not unreasonable nowadays either, if >> you're using something like DPDK. Go look at companies like 6WIND, they >> claim on an Intel Xeon CPU, being able to do in excess of 5 million PPS on >> a single core. Apparently scales as far as you can go. Granted, there's a >> lot of development work to use DPDK, but it's allegedly possible. >> >> I think most of the Procera stuff is actually done in software, I don't >> think they have any dedicated ASICs, it's all Intel hardware. They probably >> use DPDK. >> >> On 04/22/2015 10:37 AM, Josh Reynolds wrote: >> >> I think you're thinking about Saisei or whatever. >> >> Procera is done in hardware :) they also can stack their management in >> distributed deployments. >> >> On April 22, 2015 7:32:22 AM AKDT, Ken Hohhof <[email protected]> >> <[email protected]> wrote: >>> >>> Looks like I was wrong, they have some pretty big boxes. >>> >>> 600 Gbps and still all in software? >>> >>> *From:* Josh Luthman <[email protected]> >>> *Sent:* Wednesday, April 22, 2015 10:07 AM >>> *To:* [email protected] >>> *Subject:* Re: [AFMUG] Procera sold for $240M >>> >>> I was gonna say...isn't that kinda their market? >>> >>> >>> Josh Luthman >>> Office: 937-552-2340 >>> Direct: 937-552-2343 >>> 1100 Wayne St >>> Suite 1337 >>> Troy, OH 45373 >>> >>> On Wed, Apr 22, 2015 at 11:04 AM, Paul Stewart <[email protected]> >>> wrote: >>> >>>> “I’m not sure Procera has a box for the really big carriers like >>>> Comcast, AT&T, Verizon. I assume Google Fiber will design and build their >>>> own, unless they totally believe in throwing bandwidth at the problem.” >>>> >>>> >>>> >>>> http://www.proceranetworks.com/products/pl20000 >>>> >>>> >>>> >>>> >>>> >>> >>> >> >> -- >> Sent from my Android device with K-9 Mail. Please excuse my brevity. >> >> >> -- >> Simon Westlake >> Powercode - The smart choice in ISP billing and OSS >> powercode.com >> P: 920-351-1010 >> E: [email protected] >> > > -- > Sent from my Android device with K-9 Mail. Please excuse my brevity. > > > -- > Simon Westlake > Powercode - The smart choice in ISP billing and OSS > powercode.com > P: 920-351-1010 > E: [email protected] >
