That's a nice initiative and good timing, we will be interesting and
willing to contribute as well, if we go with tungsten Fabric.
We recently looked at using Contrail, the commercial version of Tungsten,
not very impressed for managing network switches to be honest.
Overcomplicated software with limited configuration schema for switches.
Although we haven't looked at the service orchestration part.
It seems to be somewhat useful when it comes to orchestrate services such
as firewall, LB, vxlan routers,...
Will also be challenging to deal with their documentation which lacked a
bit and also their high integration to OpenStack :-(
We are expecting to have a close eye on Tungsten for our network services
evolution.
Perso, I would look for another SDN, but I'm not knowledgeable in this
topic at the moment, it might be the best open source option available at
the moment.
Cheers,
On Mon, Feb 3, 2020 at 11:51 AM Simon Weller
wrote:
> All,
>
> During the 2019 CCC @ Apachecon North America, a few of us discussed the
> need for a new Software Defined Networking (SDN) integration for
> CloudStack, now that Nuage has chosen to depreciate their SDN product
> portfolio.
> I've been working closely with Sven Vogel on outlining how we might be
> able to start a Special Interest Group (SIG) to design and build an ACS
> network plugin into the Linux Foundation project Tungsten Fabric (Formally
> known as Open Contrail).
>
> Both Sven's company, EWERK and my company ENA are willing to contribute
> developers to this effort, as we feel it's important for Apache CloudStack
> to have a robust SDN option that utilizes a well known and stable open
> source SDN project and one that is community supported.
> Although there is an existing plugin for Contrail that was originally
> contributed by Juniper, it has been orphaned in the ACS code base for many
> years and to my knowledge, it’s unusable.
> Over the years, we've had a number of SDN integrations come and go. This
> has left users in the lurch and discouraged other potential companies from
> considering these options, as one has to be confident in the longevity of
> the plugin.
>
> Why Tungsten Fabric?
> Tungsten Fabric has been around for quite a while and it is now officially
> a Linux Foundation project, so it has a considerable amount of support
> behind it. It's scalable, multi-tenant, supports a number of advanced
> security features, as well a large chunk of built in components we
> currently need a Virtual Router to provide.
> It's dual stack IPV4 and IPV6 and heavily utilizes BGP and MPLS. This
> makes it ideal for those of us that maintain our own networks, as it will
> provide tight integration options and eliminate the need for complicated
> Private Gateway (PG) setups for VPCs.
> Additionally, with service stitching and EVPN capabilities, it will make
> it a lot easier for operators to support other platforms without having to
> build a dedicated plugin, or figure out how to support those network or
> security features through other L2/L3 hacks.
>
> Sven and I would like to gauge feedback from the community on this
> proposal and see whether other organizations are interested in
> participating.
>
> Thanks,
>
> Sven and Simon
>
