It is Wikibon’s belief that the trends put forth by hyperscale companies are starting to push into the enterprise to create a software-led datacenter. While SDN fits into this trend, there is an even greater opportunity to truly transform the way that the network is managed and potentially disrupt the current industry structure, which is dominated by Cisco. While customers may complain about high prices, networking has traditionally been bought on risk-avoidance and changing vendors can cause a lot of angst and internal resistance. Cisco has done a great job of keeping up with customer requirements over the last 15 years and despite industry consolidation – both acquisition and challengers in the channel – has managed to maintain a long lead, especially in L2/L3 switching. There are rumors that Cisco’s latest spin-in – Insieme Networks – will make some announcements at Cisco Live next week. Whether that happens or not, with Cumulus Networks coming out of stealth today, let’s take a look at some competing visions for where networking is heading.
With some high profile executives and investors, Cumulus Networks has been hotly talked about in Silicon Valley (especially after the Nicira acquisition) and has now emerged from stealth mode. The company is positioned as “bringing the Linux revolution to networking”. Cumulus is a software company that creates a full networking solution by leveraging “white box” switches (see my take on how commodity hardware fits with SDN). If SDN solutions can successfully simplify L2/L3 protocols and management, the underlying hardware’s primary function is to provide a reliable and scalable substrate. Following Cumulus’ analogy of Linux, the switch operating system is the platform that can be used with any standard hardware and applications can run on top of it. This connects well with Nicira’s vision of network virtualization. The next phase of IT is the creation of platforms where modern applications can be developed – such as what Pivotal is trying to do, what OpenDaylight could possibly do with SDN, and also a huge opportunity for OpenStack. Wikibon’s Dave Vellante has stated that the biggest winner of Linux was not the distributions like Red Hat, but Google, which built a whole business model around leveraging Linux. Cumulus Networks extensible, scalable and low-cost structure addresses the needs of those who look to embrace cloud and big data solutions. Similar to Nicira, in the early days, this solution makes the most sense for service providers and large enterprises whose IT deploys like an internal service provider.
The team that has brought tons of innovative solutions to Cisco including 2 spin-ins – Andiamo (FC switching) and Nuova (Nexus switching and Unified Computing System (UCS)) – has been in deep stealth on Insieme Networks. While the company is doing a very good job of keeping plans under wraps, while I have not had a briefing, I’ve heard a few (unconfirmed) things and have some questions that I would ask when they start talking publicly. First of all, Cisco is not building a storage solution. Will Insieme have some integration with storage, probably – Andiamo had FC and Nuova has FCoE, so working with storage networking protocols and even integrating with management is important. Cisco’s data center strategy includes lots of partnerships with storage – NetApp FlexPod, VCE Vblock and EMC VSPEX drive a significant percentage of UCS sales and at Cisco Live a bunch of other storage partners including HDS, Tegile, Nimble Storage and Nexenta are exhibiting. There is always industry chatter about Cisco making a storage acquisition; Cisco needs to have plenty of flash options, but is best served to continuing to partner in the complex and highly fragmented storage market. So back to Inisieme, core to Cisco’s switching is how it can continue to differentiate from the merchant silicon crowd. Mario Mazzola & team have plenty of software engineers, but at the core, they focus on creating ASICs. I would be shocked if the Insieme message does not include how Cisco will have tons of speeds and feeds to compete with the latest Arista and others who are trying to beat Cisco in the 40Gb and 100Gb Ethernet markets. There should be an SDN play for Insieme, of course Cisco is already making progress on being the default OpenFlow controller thanks to OpenDaylight. If Insieme was an independent start-up launching a new switching architecture, I do not think that it would have much chance to penetrate the market. Cisco has such an advantage with sales, marketing and customer install base that it is hard to bet on them not being able to find a market for a new solution.
As CIOs and IT organizations examine the transforming networking industry, it is not an simple decision of “open vs. closed” or “software vs. hardware”, but rather how IT can drive value for the business. Enterprise organizations look for solutions that can help save time and therefore services and ecosystems are needed for the open source solutions. There is general consensus that networking will undergo some seismic shifts over the next 5-10 years, but the winners are still to be determined.
For more on Cumulus Networks, see James Hamilton’s blog post.
For more on Insieme Networks, see Shamus McGillicuddy’s article Leaked: Insieme Networks is building massive fabric controller, Jim Duffy’s Arista/Insieme 100Gb post and Brad Casemore’s older article Cisco’s SDN Strategy: Meet the New Boss, Same as the Old Boss.