[AusNOG] Open Networking

Wed Jul 6 10:19:01 EST 2016

On 6 July 2016 at 01:49, Nik Geyer <nik at neko.id.au> wrote:

> I thought Quagga now had LDP support (LDPd port from OpenBSD) and Cumulus
> was working on the Quagga -> Zebra -> Kernel LFIB code as we speak, and it
> should be available "real soon now".
>

I think this is one of the things that has saddened us most about Cumulus -
we've been told that this was coming for around 2 years now - and "real
soon now" has been the same story. VRF/namespace support finally came into
play in Cumulus 3 which we haven't gone to test yet… and that's a good
"first run". My understanding is they have a number of SP customers in the
US that are keen for this though, so I'm sure they will get there.

> The argument around why buy a $40k router when a $10k switch will do the
> job is a valid one, just make sure you validate all your technical
> requirements properly. For example, the way the UFT's are carved up on
> Trident can be odd at times, there are LFIB table size limits (that are
> different for ingress or transit labels), and so on. If it ticks all the
> boxes then go for it. Also if you need long distance optics or coherent
> optics, you'll be back to using a router ;)
>
>
Certainly there is a lot to be said for being able to get a $10k switch to
do the job of a $40k router - and I think we'll see more in this space.

I also understand there are some people out there looking at using OpenFlow
and x86 based controllers to make packet decisions and traffic optimisation
so you essentially have a switch acting as an openflow endpoint, receiving
it's routing direction and intelligence from a faster/more beefy hardware
box. There are some compromises that would have to be made in terms of
packet direction, but given the increase in x86 routing performance this
could really strike a balance between high performance and low cost
routing. (In some senses it's just a bigger abstraction of the
control/route plane design we've bee running for years…)

All that said - Lincoln's comments about testing are absolutely necessary.
We've found a number of issues with the few vendors we've run with, and
testing to know what they are before we flow into production is absolutely
necessary (and… now even more realistic given the higher availability of
virtualised routing products… just whose bug are you finding :-) )

Thanks,
Andrew

> Sent from my iPhone
>
> On 4 Jul 2016, at 12:51 PM, Ben Hohnke <settra+ausnog at gmail.com> wrote:
>
> We're looking into it, as there is a large cost saving to be had.
> Unfortunately, as we have a high reliance on MPLS for our network, our
> options are limited. Cumulus are adding in MPLS support soon, but its only
> BGP label distribution, which is not very handy in a large network. I've
> been talking with them, however, and am considering a limted trial in our
> network for some L2 stuff, in conjunction with some Agema Trident2 switches
> we have gotten in.
>
> We're testing switch and virtual router software from Ip Infusion, but I
> wouldn't really call it "open" - the routing stack is based on Zebra I
> think, but everything is closed source.
>
> We use VyOS a little where we don't need MPLS, and it does the job.
>
> I've done some testing using OpenBSD and LDPd with OSPFd, but I was seeing
> some weird behaviour when I fed it a full LDP and OSPF table from our
> network.
> We are a mainly Cisco and Mikrotik shop, however this is starting to
> change as newer players enter the market, and are able to beat the bigger
> players on price.
>
>
> Longer term, I expect our core to contain little to no Cisco / $bigvendor,
> mainly because we're at a scaling point where the $ vs performance is
> heavily weighted against any of the big players. Why get a $40k router that
> can do 20-30Gb/s of forwarding when you can get a  $10k L3 switch that can
> do it all at line rate?
>
>
>
> On Mon, Jul 4, 2016 at 12:35 PM Simon Attwell <simon at attwell.net> wrote:
>
>> Hi All,
>>
>> Just curious how many of you have deployed / are deploying / Open
>> Networking in production environments.
>> I'm interested to see if ON is making its way down to the edge (1Gbps
>> PoE/PoE+) or if it's mainly being used at the distribution / core layers or
>> at the service provider level where there's little end device connectivity
>> and it's more about moving the packets around.
>>
>> Comments on hardware choice / stability / longevity / MTBF / support, are
>> also appreciated.
>>
>> From a Cumulus perspective it looks like 1 Gbps - 100Gbps is where things
>> are focused.
>> Nothing with PoE/PoE+ support so it looks like at the moment we're only
>> talking about datacenter switching.
>>
>> What I don't see deployed today is a lot of technology mix, especially in
>> switching. Customers have a preference and for support / interop / personal
>> reasons tend to stick with a single vendor for switching.
>> In the past this has made sense as switches did not always play well with
>> others.
>>
>> I'm wondering what you all think the 3 - 5 year picture looks like.
>>
>> I suspect it looks a lot like the current virtualization market. A few
>> major players with custom software built on open source foundations, being
>> hardware agnostic and the holdouts trying to ignore the fact that the
>> industry is fundamentally changing.
>>
>> - Simon
>>
>>
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> AusNOG mailing list
>> AusNOG at lists.ausnog.net
>> http://lists.ausnog.net/mailman/listinfo/ausnog
>>
> _______________________________________________
> AusNOG mailing list
> AusNOG at lists.ausnog.net
> http://lists.ausnog.net/mailman/listinfo/ausnog
>
>
> _______________________________________________
> AusNOG mailing list
> AusNOG at lists.ausnog.net
> http://lists.ausnog.net/mailman/listinfo/ausnog
>
>

-- 
*Andrew Yager, CEO* *(BCompSc, JNCIS-SP, MACS (Snr) CP)*
Real World Technology Solutions - IT People you can trust
Voice | Data | IT Procurement | Managed IT
rwts.com.au | 1300 798 718

*Real World is a Dell Premier Partner*

This document should be read only by those persons to whom it is addressed
and its content is not intended for use by any other persons. If you have
received this message in error, please notify us immediately. Please also
destroy and delete the message from your computer. Any unauthorised form of
reproduction of this message is strictly prohibited. We are not liable for
the proper and complete transmission of the information contained in this
communication, nor for any delay in its receipt. Please consider the
environment before printing this e-mail.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ausnog.net/pipermail/ausnog/attachments/20160706/eaf67223/attachment.html>