Daily Archives: March 12, 2012

Arista’s Adaptable Approach to SDN

In an earlier post regarding Arista Networks’ march toward an IPO, I wrote that I would provide an overview of the company’s positioning on software-defined networking (SDN), which now follows. I think the subject is worth exploring given the buzz generated both by the IPO-bound Arista, with its notable market successes in high-frequency trading and other application environments requiring low-latency switching, and by SDN itself.

Last fall, when OpenFlow fever reached a boiling point, Arista Networks’ CEO Jayshree Ullal pointed out that it was just one mechanism of many that could be leveraged in the service of SDN. Among the others, she opined, were existing command-line interfaces (CLIs), Simple Network Management Protocol (SNMP), Extensible Messaging and Presence Protocol (XMPP), Network Configuration Protocol (NETCONF), OpenStack (with its Quantum project), as well as APIs in VMware’s vSphere virtualization software.

The Four Pillars

On the larger SDN canvas, Arista has propounded its “four pillars” of software-defined cloud networking (SDCN). You can read about Arista’s “four pillars” in a blog post written late last year by Ullal or in a white paper that can be found on Arista’s website. In both, the four pillars are identified as follows:

Pillar 1. Single Point of Management, which Arista believes can be achieved through layering atop the traditional control plane and data path of a cloud network and through coordinating configurations across multiple otherwise-independent switches. Arista says no fabric technology is required, and it says its CloudVision is up to the challenge.

Pillar 2: Single-image L2/3 Control Plane.  Here, Arista believes “standards-based L2/L3 IETF control-plane specifications plus OpenFlow options (without hype) can be a promising open augmentation for providing single image control planes in the future.”

Pillar 3. Multi-path Active-Active Data Path. The company prescribes scaling cloud networking across multiple chassis with Multi-Chassis Link Aggregation Group (MLAG) at L2 Equal Cost Multi-pathing (ECMP) at L3.

Pillar 4. Network-Wide Virtualization. Regarding this last pillar, the company says it makes sense to provision the entire network to handle any application seamlessly and so that the economics of virtualization can be properly leveraged “using controllers from VMware and their new paradigm for VMWare’s VXLANS or Open Virtualization Switching (OVS) controllers in the future.”

Best of Both Worlds?

As has been above (and in earlier posts), software-defined networking can be implemented in more than one fashion. Some networking vendors — typically industry mainstays with large installed bases of customers and firmly established business models predicated on hardware ASICs, proprietary protocols, and relatively high margins — will opt for an SDN vision that features a distributed control plane. Not for them the dramatic shift to logically centralized server-based controllers, designed to subsume networking within a computing paradigm. To the traditional networking vendor, that road looks treacherous and leads to a diminution of the status and margins associated with the beloved switch.

As neither a raw SDN startup nor a legacy networking company, Arista takes a flexible position on how SDNs can be realized. The company says customers can implement SDNs by using controllers or by using distributed-control mechanisms. Ideally, according to Arista, both means should be employed for comprehensive SDN capabilities. A presentation available online explains the company’s position on this best-of-both-worlds approach to the control plane.

Finally, it probably comes as no surprise that Arista prescribes its own Linux-based Extensible Operating System (EOS) as the appropriate software foundation for its four pillars and for cloud networking in general. It also believes that “good old fashioned Ethernet scaling from 10 gigabits to 40 gigabits to 100 Gigabits and even terabits with well-defined standards and protocols for L2/L3 is the optimal approach.”

In view of the media blitz undertaken by Arista founders Andreas Bechtolsheim and David Cheriton late last year, we should expect the company’s next generation of switches to deliver as much bandwidth as Ethernet and merchant silicon will allow.