Matthew Palmer’s discussion of Cisco’s acquisition of Meraki briefly addresses whether one can consider Meraki an “SDN” company. The question raises some important aspects of SDN and the value proposition that I think are worth poking at a little more. To be fair, I’m probably less theological about the definition of SDN, but to me three important aspects are (1) the modernization of the control plane of a network, (2) logical networks built as an overlay on a transport network and (3) the migration of some “network” function into conventional software (e.g., a VM running on a virtualized server). From this perspective Meraki (and Pareto Networks and Aerohive) are very SDN like. Each builds a controller that is external to the network, builds a logical network as an overlay over the Internet and adds functionality implemented as conventional software.
A key aspect of each is the delivery of some elements network management functionality as SaaS. The value proposition of SaaS is the dramatically improved development rate and support cost if you deliver software in the form of a multi-tenant, single instance shared service (the most visible example is of course Salesforce.com). If all users run on a single shared instance of the software then everyone has exactly the same software and hardware configuration and encounters exactly the same bugs (support is greatly simplified). Conversely, new features and bug fixes are quickly available to everyone. Where the remote software functionality makes sense, it’s a better way of getting benefit of the software, but of course there are various devils in the details.
One important detail, especially with respect to SDN, is latency. A server that is LAN connected is a lot more responsive than an Internet-connected SaaS service. Given that controller access latency is already an issue in a LAN connected OpenFlow controller design, it doesn’t make obvious sense to think of building a data center OF SDN controller as a SaaS service. On the other hand, for Pareto’s initial product offering (simplified branch office communication) the SaaS benefits far outweighed any issue of latency. And as Zscaler or CloudFlare demonstrate, you can do much to improve Cloud service latency when it’s important.
Over time, doing much of enterprise network management as SaaS seems to me a really good idea, and Meraki is a good step in that direction, and as Matt said, makes sense as a product offering for Cisco. It’s as much SDN as anything in my book, but admittedly I’m not viewed as very SDN politically correct.