Article written by Scott Thompson of FBR Capital Markets.

References provided by Alan J Weissberger of IEEE ComSoc

OCS Report:

On January 28 and 29, the Open Compute Summit held its annual gathering where hyperscale and large enterprises show cased the creative and cutting-edge work they have pioneered over the past year in a quest to drive costs and complications from the IT landscape. We attended the conference and present some of the more important points below.

    * One of the messages with the most impact at Open Compute this year was delivered by Facebook CEO Mark Zuckerberg. He suggested that one of the top strategic priorities at Facebook is to find a way to deliver Internet service to a larger portion of the world’s population.  Mr. Zuckerberg stressed that emerging markets (two-thirds of the planet’s population) often do not have the same access to the Internet as does the developed world. Facebook believes that finding a way to provide affordable access to the Internet for those markets represents one of Facebook’s largest growth opportunities. He added that the costs Facebook presently incurs to deploy networks that run its services would need to decline to one-tenth of current rates to make this goal financially achievable. He also added that Facebook has a plan to deliver those types of network efficiencies over the next five years and that the Open Compute Project is an essential cornerstone of that effort.

    * We found a chorus of network pioneers at the event that agreed with Mr. Zuckerberg’s assessment. The networking and compute process, when combined in ways that webscale operations have been perfecting for several years, is likely to deliver significant savings, compared to traditional methods and architectures.

Intel is introducing reference designs for “rack-scale architecture” that, by 2015-2016, should dramatically change the way the world processes data and manages compute resources and networks both together.

During a panel hosted by Facebook’s Director of Infrastructure Najam Ahmad, a large number of panelists appeared to confirm that there is a clear and new reference design for how the IT community is beginning to measure efficient datacenter infrastructure.

Martin Casado, PhD & CTO of networking at VMWare, and J.R. Rivers of Cumulus Networks both seemed to agree that the new distributed datacenter model used by many hyperscale networks has achieved new levels of efficiency. Mr. Casado suggested these reference designs, across both enterprise and service verticals, are quickly becoming the de facto standard by which new and more efficient networking technologies can be deployed. He suggested that we may be nearly past the technical barriers required for achieving hypserscale efficiencies and that the primary constraint keeping enterprises and service providers from adopting similar technologies is now cultural. Casado believes that educating those who design and deploy IT systems is now the most effective way to bring about wholesale change across the networking landscape.

    * The Open Compute Project (OCP) appears focused on this challenge in 2014. While the pure pace of innovative technology on display at this year’s Open Compute Summit appeared to have slowed, relative to the last one, the mechanisms and framework needed to effect and propagate change appear to be improving dramatically.

The OCP appears to be rapidly gaining the support of the vendor community, bringing it closer to an ecosystem for the acquisition of bespoke networking and compute equipment. From a logical  perspective, we continue to have reservations that an “open community” will have the scale and purchasing power to effect change across vendors that might prefer other solutions, but we remain impressed by the scale, scope, and momentum that the OCP has created over the past few years. Whether it proves to be long term or not, Open Compute and its member organizations are changing the face of the IT landscape.

References:

1. SDN Central’s OCP Summit Report:

http://tinyurl.com/m9442bo

2. OCP-Network Group Charter & more

http://files.opencompute.org/oc/public.php?service=files&t=437680b2416f9…

3.  Broadcom’s Leaf & Spine Network Switch Spec

http://www.opencompute.org/assets/Uploads/Open-Compute-Project-BRCM-Open…

4.  Intel’s 10/40 Gigabit Ethernet Rack Mountable Switch Standard, by Michael Miller

Contact: [email protected]  for a soft copy (pdf)