[dpdk-dev] dpdk vs netmap

[dpdk-dev] dpdk vs netmap

[Bob Chen]

What's the advantage of DPDK over another accelaration framework called netmap?

Re: [dpdk-dev] dpdk vs netmap

[François-Frédéric Ozog]

Well done! 

I guess that's the shortest question on the list, and probably the one
that's going to trigger the largest discussion.

A few months, ago I had to answer it for a customer. And here is my
understanding:

- DPDK is also a high performance multi-core application framework. You take
out the packet framework and DPDK has still value because it allows you to
get the most of a multi-core environment. In other words, if you want to
develop say an accelerated web server, you have better chances to achieve
the best results with DPDK because it does not stop providing performance
helpers at the "packet pump" level: threading model, high performance memory
management, high performance locking in userland...

- DPDK programming requires more effort at the beginning because of this
multi-core environment

- DPDK is now at the center of an informal eco-system. Not too long from
now, I assume you will have a "DPDK-store" from which you will obtain either
freely or with a fee a lot of additional components: TCP/IP stack...

Now if you want detailed comparison on the performance aspects.... Here is
my view: on a simple L2/L3 packet forwarding "proof of concept" with 2 ports
you will find them equal: the CPU does not do enough. So you need to model
your application and try it with at least 2 ports on both environments. And
you probably want to put it at serious test with 8, 16 or even 22 ports.
Then you'll get your answer.

Bet's are open. I put a beer on DPDK ;-)

 
François-Frédéric


> -----Message d'origine-----
> De : dev [mailto:dev-bounces@dpdk.org] De la part de Bob Chen
> Envoyé : lundi 10 mars 2014 06:49
> À : dev
> Objet : [dpdk-dev] dpdk vs netmap
> 
> What's the advantage of DPDK over another accelaration framework called
> netmap?

Re: [dpdk-dev] dpdk vs netmap

[Stephen Hemminger]

This is talk I did a couple of years ago.

http://www.slideshare.net/shemminger/uio-final