Virtualization and Cloud executives share their predictions for 2013. Read them in this VMblog.com series exclusive.
Contributed article by Steve Riley, Technical Director in the Office of the CTO at Riverbed Technology
SDN jumps from hype to reality by addressing its biggest problem
In 2013, the software-defined network (SDN) conversation will shift
away from moving bits around the lower layers of the network stack to managing
complexity at the upper layers. This will help to create a bridge between the
vision of what network virtualization is supposed to deliver and what is needed
today to make that happen: network consistency.
For the past year, network pundits obsessed about the transition
to network virtualization and debated the merits of one tunneling protocol over
another. While such nuts-and-bolts thinking is necessary, we must continue to remember the
goal: to solve difficult and challenging networking problems. Server
virtualization-converting servers from silicon to software-provides useful
operational capabilities. In addition, provisioning is nearly instantaneous, and
snapshots and rollbacks are now simple scripts. Effective network
virtualization-where the network is similarly a software entity-must provide
similarly useful operational capabilities. When the control plane is decoupled
from the physical network fabric and reconstituted as a collection of logical
abstractions, many of the usual constraints (topology, connectivity, limited
state awareness, the pain of renumbering) fade away.
Now that the SDN die has been cast, what will matter next is where
the innovation lies.
As I described in a recent article, with SDN, the control plane is built with software
running on ordinary server computers (while the forwarding plane continues to
exist in hardware). One of the challenges of this is network consistency. Ensuring
consistency requires a lot of intelligence because virtual networks need to know
everything about all nodes and ports, make sure that all of the paths are
consistent whenever it comes time to plumb a new connection, and do all of this
battle is brewing over where these brains should go in an SDN, with various
interests jockeying for "enhanced semi-standard" control plane implementations
(sometimes implemented in dedicated hardware components!). Business wants the
flexibility to program SDNs and, frankly, the market will determine how this
will be offered - open-source or proprietary - it really doesn't matter.
my belief is that software controllers are the key to a programmable
infrastructure, and virtual networks constructed of such software will turn the
cloud to the reality that we've wanted. It is hard work and scary-but necessary.
means that the internal application and network teams will have to collaborate
and play in the same sandbox because application performance is no longer isolated from the network. These
teams will need to work together in order to ensure the performance, control,
and security of applications.
In addition, a self-service
network development model will evolve with fully automated provisioning of cloud
environments that enable enterprises to visualize and
control their networks with their own tools. In essence, they'll "program"
their network with their own parameters. It is the ultimate in flexibility and
innovation as more enterprises realize that they can rapidly and dynamically
carve up their network and data center as needed. By spooling up and down entire
functional ensembles of servers and networking to meet changing business
demands, SDN will offer up a level of flexibility and nimbleness that is not
available with traditional networks and data centers.
This is the SDN conversation I look forward to having in 2013.
About the Author
Riley is a Technical Director in the Office of the CTO at Riverbed Technology.
His specialties include the performance and security aspects of enterprise and
cloud computing. Steve has a long career of public speaking, having participated
in hundreds of events around the world. He is co-author of Protect Your Windows Network, contributed a chapter to Auditing Cloud Computing, has published
numerous articles, and conducted technical reviews of several data networking
and telecommunications books. At Riverbed, Steve concentrates on
high-performance architectures that span multiple clouds, public and private.
Before Steve joined Riverbed, he was cloud security strategist at Amazon Web
Services and a security consultant and advisor at Microsoft; in both
capacities, he developed patterns and practices for secure deployments and
operations. Steve is a member of the Kubuntu Team (which
maintains Ubuntu's KDE-flavored distribution) and is a global moderator of its
community forum. Besides lurking in the Internet's dark alleys and secret
passages, he enjoys freely sharing his opinions about the intersection of
technology and culture. Contact him at email@example.com.