Today, Cumulus Networks came out of stealth mode. Numurous blogs, for example this one, are calling Cumulus the Cisco-killer. Of course this is extremely flattering to Cisco. It establishes your position as the market-leader when everybody tries to ‘kill’ you. I have seen the same, more intensive, with Apple, where every other smartphone of tablet was being called the iOS-killer. (do you notice the name similarity?)
But is it really a Cisco killer? Cumulus is basically putting Linux on commodity hardware. For someone who puts OpenWRT on every possible SOHO-device, this is the next step. But I am not sure if this will be the thing large-scale networks are looking for. Datacenters however, and especially the ones with Openstack, will embrace Cumulus. The network can now be managed as a server and integrate with all the current orchestration tools, like Puppet and Chef.
Network engineers know that commodity hardware is desirable, but it cannot mean multiple network operating systems. Software is the major operational issue in networking, and managing different systems ultimately means operational costs that far exceed the purchase price.
While I see his point, a multi-vendor strategy makes it possible to drive down the cost on the Capex side. A network abstraction layer with orchestration possibilities makes sure you won’t get hit on the Opex side. Making the case for this kind of cheaper hardware.
But the killer comment comes from a competitor:
Unfortunately for them, that point brings us directly to that ever-so-slightly-sticky issue of embedded control vs. logically centralized control in a network, which is at the heart of the Open Networking Foundation discussion. Simply put, Cumulus Linux is the polar opposite of a Cisco Killer — it’s a full-on endorsement of the traditional Cisco-style architecture, just at a lower price point. Their L2/L3 model is the Cisco model.
Let’s be clear, the traditional L2/L3 model is not dead, yet. The promise of SDN still has to proof it’s self, but being Linux based, in theory there is the possibility to evolve into a full SDN-architecture.
The least we have won with the introduction of Cumulus is the separation of hardware and software. Although it is yet unclear to me if the system is built on open source or proprietary binary drivers. A thing felt as problematic with the current hardware vendors.
While the hardware compatibility list still has to grow, I see a future for Cumulus. It’s going to be a niche-market, but within that niche the need for such a vendor is extremely high. But to call it a Cisco-killer? No way, iOS maybe ;-)