In 2014, Arista was telling me that SDN was “Still Don’t kNow” and “Still Does Nothing”. In June 2015, Arista releases its CloudVision™ SDN controller for cloud networks.
So Why Now ?
Arista was reluctant to engage with SDN in 2014. With its public listing on the stock exchange, it needed business to run smooth. Plus it was a messy time as Cisco chose to make its SDN hardware-centric while VMware NSX remains resolutely software-focused. OpenStack has been evolving and changing direction, which undermines confidence. Plus lots of startups were competing for the limited number of early adopters.
In mid-2015, SDN is a requirement for the data centre where the convention is to have a 10-year design strategy and product lifecycle. More importantly, SDN platforms move the sales conversation away from the switch hardware to the platform strategy for network automation and orchestration.
Cloud Providers and Growth
Arista has been successful in selling to cloud providers, notably Microsoft, but those mega-providers can now easily build their own network switches that meet their own requirements. Amazon, Facebook, and Google have been public about building their own software for whitebox switches, and more recently building their own hardware.
It’s not hard to predict that Arista needs to grow fast. And the Enterprise can be a strong market for Arista as Juniper loses focus and Cisco has several SDN strategies in play. Private cloud networking is going to be a big deal in a couple of years and Arista needs to make sure it has a product there.
Cloudvision
Arista suggests network automation falls into three broad categories – DIY, DevOps, and Turnkey.

In order to help all of these customers, EOS CloudVision™ is a centralised controller for network configuration and control for all Arista EOS devices. This should be familiar to most people. The actual product ships as a software VM that can be installed on your platform of choice.

And it plays nice with everyone by integrating with other vendors’ tools. (Note that all the vendors do this; it’s a great way to remove customers objections and cross-sell into new accounts.)

I like the way Arista has painted the controller landscape here and highlighted that there are many approaches to SDN controllers.

Best Feature – Network Rollback
Arista offers a network rollback capability that uses rolling snapshots to rollback on demand. The feature is neat but it highlights the limitations of switch configuration today where the internal architectures are not flow programmed. Until this is possible, the use of device configuration remains a serious operational risk during configuration changes. A rollback to regular snapshots is the best we have with the current generation of networking protocols.

Some Observations
- CloudVision™ is shipping today. No waiting around or working your vendor to get onto a shortlist for special access.
- Subscription pricing is available for people who don’t have enough money to buy it.
- I get the sense that Arista is working closely with VMware NSX to become a preferred partner for the physical network. This makes sense when Cisco is aggressively competing in the same markets. The problem is that VMware is only interested in the data centre and not in Campus or WAN, where Arista needs to be in a few years.
In my world view, Arista had hoped other vendors would deliver SDN controllers and applications that would use Arista hardware. However, standard bodies have failed miserably (again) to actually produce standards, and OpenStack has changed the network landscape in the data centre, where whitebox and whitebrand could become a threat to Arista’s business. Hence we have another SDN controller platform that abstracts a vendor’s hardware platform from the SDN applications.
Strengths
- Arista uses EOS software from its switches as the base for the controller.
- Arista has been working with cloud providers on network automation for a long time; it’s likely they have learned something from this.
- Arista’s focus on software and product quality is a good sign that ClearVision™ might work as promised.
- The company had plenty of time to see missteps from competitors.
- It creates a single point of configuration and control for many devices and includes visibility and analytics.
- It has a single API point for northbound applications.
- It has unique enhancements for streaming telemetry data to applications (instead of polling).
Weaknesses
- Arista is late to the SDN market and will need to capture customer attention.
- Arista CloudVision™ needs partnerships and integration to be successful. How many “partnerships” can other vendors reasonably sustain when simply being part of the OpenStack strains budgets & resources?