Virtualization and Cloud executives share their predictions for 2014. Read them in this VMblog.com series exclusive.
Contributed article by Dinesh Dutt, Chief Scientist at Cumulus Networks
2014: The Beginning of a Truly Open Networking World
This was the year
the white box genie got out. The year when most companies vied to announce how
open and Linux-focused they are. The year when software really defined
networking, if not always the way many predicted.
They say revolution
usually begins at the fringe and this proved true with networking too. It began with what are considered niche or fringe customers such as Google, Microsoft and
Amazon. If Luke Skywalker and the rebel alliance on the fringe are lucky, 2014
will see the accelerated breakup of what is now a vertically integrated
networking world, leaving behind a noticeably weaker and smaller shell that will
continue to erode during the coming years. If they're not lucky, then 2014 will
be the year the empire strikes back.
In order to see a
truly open networking world, opposite of what the empire has positioned in the
past as the status quo, we need to be able to choose at multiple levels: the
silicon, the hardware, the software and the applications that run on top.
The merchant
silicon industry is already beginning to see some competition with seasoned
firms like Intel and Mellanox being joined by startups like Centec Networks to provide
alternatives to Broadcom. The coming year will see the arrival of open hardware
based on some of these boxes.
The hardware
industry too is being seeded with a lot of players, from the more established to
relative newcomers. The availability of network OSes to power these open
hardware boxes, and the initiatives like Facebook's Open Compute Project, ease
the entry of companies into this segment. This market will continue to mature
rapidly in the coming year, bringing about more CPU choices and more
standardized hardware allowing zero touch provisioning of bare metal
switches.
I also predict
applications will be the new battleground. When software truly starts to define
networking, and vendor lock-in will move to software. This is especially true
for applications in the cloud. To counter that, open source applications like
OpenStack will have to make L3-based network overlays first-class citizens.
Network OS is and
will continue to be a key player in this ecosystem. It is the glue that allows
for open routers and switches while providing a stable base for applications to
run on. This is a fundamental reason to keep the network OS distributed and
autonomous, rather than making it a puppet in the hands of a centralized, locked
in controller. A true Linux network OS model is crucial in the year ahead, as
with this, layers above and below it can change while the management of the
infrastructure truly stays open and easy. And, a mature platform built by a
vibrant, innovative and responsive community ensures that we can all truly
benefit by working on the problems collectively. The year ahead will be crucial
for network OS.
Open networking is
not a spectator sport. As consumers and end beneficiaries, we can affect the
final outcome. If we're deceived by merely a veneer of openness, then the years
ahead will reveal that the new networking is just as difficult, inflexible and
slow as the old one was. It just won't be immediately apparent. This will be the
year we need to work together.
##
About the Author
Dinesh Dutt is
Chief Scientist at Cumulus Networks. He was a Fellow at Cisco Systems before
this. He spent almost 15 years at Cisco Systems working on various key
technologies such as improving enterprise L2 networks, SAN networks, design of
high end switches and most recently, data center networks. He's a co-author of
the TRILL and VxLAN protocols.