Re: [dpdk-dev] KNI and memzones

["Zhou, Danny" <danny.zhou@intel.com>]

It looks like a typical network middle box usage with IDS/IPS/DPI sort of functionalities.  Good enough performance rather than line-rate performance should be ok for this case, and multi-threaded KNI(multiple software rx/tx queues are established between DPDK and a single vEth netdev with multiple kernel threads affinities to several lcores) should fit, with linear performance scaling if you can allocate multiple lcores to achieve satisfied throughput for relatively big packets.

Since NIC control is still in DPDK’ PMD for this case, bifurcated driver does not fit, unless you only use DPDK to rx/tx packets in your box.

From: Jay Rolette [mailto:rolette@infiniteio.com]
Sent: Wednesday, September 24, 2014 2:53 AM
To: Zhou, Danny
Cc: Marc Sune; <dev@dpdk.org>; dev-team@bisdn.de
Subject: Re: [dpdk-dev] KNI and memzones

I can't discuss product details openly yet, but I'm happy to have a detailed discussion under NDA with Intel. In fact, we had an early NDA discussion with Intel about it a few months ago.

That said, the use case isn't tied so closely to my product that I can't describe it in general terms...

Imagine a box that installs in your network as a transparent bump-in-the-wire. Traffic comes in port 1 and is processed by our DPDK-based engine, then the packets are forwarded out port 2, where they head to their original destination. From a network topology point of view, the box is mostly invisible.

Same process applies for traffic going the other way (RX on port 2, special-sauce processing in DPDK app, TX on port 1).

If you are familiar with network security products, this is very much how IPS devices work.

Where KNI comes into play is for several user-space apps that need to use the normal network stack (sockets) to communicate over the _same_ ports used on the main data path. We use KNI to create a virtual port with an IP address overlaid on the "invisible" data path ports.

This isn't just for control traffic. It's obviously not line-rate processing, but we need to get all the bandwidth we can out of it.

Let me know if that makes sense or if I need to clarify some things. If you'd rather continue this as an NDA discussion, just shoot me an email directly.

Regards,
Jay



On Tue, Sep 23, 2014 at 11:38 AM, Zhou, Danny <danny.zhou@intel.com<mailto:danny.zhou@intel.com>> wrote:

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org<mailto:dev-bounces@dpdk.org>] On Behalf Of Jay Rolette
> Sent: Tuesday, September 23, 2014 8:39 PM
> To: Marc Sune
> Cc: <dev@dpdk.org<mailto:dev@dpdk.org>>; dev-team@bisdn.de<mailto:dev-team@bisdn.de>
> Subject: Re: [dpdk-dev] KNI and memzones
>
> *> p.s. Lately someone involved with DPDK said KNI would be deprecated in
> future DPDK releases; I haven't read or listen to this before, is this
> true? What would be the natural replacement then?*
>
> KNI is a non-trivial part of the product I'm in the process of building.
> I'd appreciate someone "in the know" addressing this one please. Are there
> specific roadmap plans relative to KNI that we need to be aware of?
>

KNI and multi-threaded KNI has several limitation:
1) Flow classification and packet distribution are done both software, specifically KNI user space library, at the cost of CPU cycles.
2) Low performance, skb creation/free and packetscopy between skb and mbuf kills performance significantly.
3) Dedicate cores in user space and kernel space responsible for rx/tx packets between DPDK App and KNI device, it seems to me waste too many core resources.
4) GPL license jail as KNI sits in kernel.

We actually have a bifurcated driver prototype that meets both high performance and upstreamable requirement, which is treated as alternative solution of KNI. The idea is to
leverage NIC' flow director capability to bifurcate data plane packets to DPDK and keep control plane packets or whatever packets need to go through kernel' TCP/IP stack remains
being processed in kernel(NIC driver + stack). Basically, kernel NIC driver and DPDK co-exists to driver a same NIC device, but manipulate different rx/tx queue pairs. Though there is some
tough consistent NIC control issue which needs to be resolved and upstreamed to kernel, which I do not want to expose details at the moment.

IMHO, KNI should NOT be removed unless there is a really good user space, open-source and socket backward-compatible() TCP/IP stack which should not become true very soon.
The bifurcated driver approach could certainly replace KNI for some use cases where DPDK does not own the NIC control.

Do you mind share your KNI use case in more details to help determine whether bifurcate driver could help with?

> Regards,
> Jay
>
> On Tue, Sep 23, 2014 at 4:27 AM, Marc Sune <marc.sune@bisdn.de<mailto:marc.sune@bisdn.de>> wrote:
>
> > Hi all,
> >
> > So we are having some problems with KNI. In short, we have a DPDK
> > application that creates KNI interfaces and destroys them during its
> > lifecycle and connecting them to DOCKER containers. Interfaces may
> > eventually be even named the same (see below).
> >
> > We were wondering why even calling rte_kni_release() the hugepages memory
> > was rapidly being exhausted, and we also realised even after destruction,
> > you cannot use the same name for the interface.
> >
> > After close inspection of the rte_kni lib we think the core issue and is
> > mostly a design issue. rte_kni_alloc ends up calling kni_memzone_reserve()
> > that calls at the end rte_memzone_reserve() which cannot be unreserved by
> > rte_kni_relese() (by design of memzones). The exhaustion is rapid due to
> > the number of FIFOs created (6).
> >
> > If this would be right, we would propose and try to provide a patch as
> > follows:
> >
> > * Create a new rte_kni_init(unsigned int max_knis);
> >
> > This would preallocate all the FIFO rings(TX, RX, ALLOC, FREE, Request
> > and  Response)*max_knis by calling kni_memzone_reserve(), and store them in
> > a kni_fifo_pool. This should only be called once by DPDK applications at
> > bootstrapping time.
> >
> > * rte_kni_allocate would just use one of the kni_fifo_pool (one => meaning
> > a a set of 6 FIFOs making a single slot)
> > * rte_kni_release would return to the pool.
> >
> > This should solve both issues. We would base the patch on 1.7.2.
> >
> > Thoughts?
> > marc
> >
> > p.s. Lately someone involved with DPDK said KNI would be deprecated in
> > future DPDK releases; I haven't read or listen to this before, is this
> > true? What would be the natural replacement then?
> >