Open source and the network's role in the cloud
Does commitment of major networking vendors to join and contribute to Open Stack community signal role open-source project will play in defining cloud networking standards?
The announcement of the latest release from open-source cloud-management software project OpenStack is remarkable in many ways. The rapidly growing OpenStack community is gaining ground on a mature platform--this release adds image management and support for unlimited object sizes in its object storage service software--and there were a number of new IT vendors added to the list of supporters.
ZDNet UK covered the basics of the announcement, so I won't pick it apart here. Rather, I want to focus on one of the most interesting aspects of many of the vendors announcing their participation with this release.
Namely, several of them are networking vendors, including my employer Cisco Systems, Extreme Networks and--according to Stephen Spector, the OpenStack communications manager, soon after the OpenStack press release went out--Arista Networks. These names are added to those of already active participants, such as Citrix and Dell, that bring their own networking technologies and perspectives to the table.
(Disclosure: I am associated with Cisco's OpenStack participation, but I don't use this blog to promote my day job. I'm not here to tout Cisco. These opinions are mine and mine alone.)
What is important here is that the networking industry is showing up at all, and what that means to the future of networking in cloud computing. To me, the way the network's relevance will be defined--or at least standardized--will be through open-source efforts. And OpenStack may be the effort where the industry comes together and hammers out one or more solutions.
How will cloud services and applications utilize network capabilities such as routing, bandwidth management, VPN creation and configuration, network segmentation and isolation, and so on? And, given that most developers could care less about networking, or at least would like to be able to ignore it if they can, how do you present network concepts in a form that is meaningful for application development and operations? (Hint: do we have the right network abstractions?)
Furthermore, how will OpenStack itself take advantage of networking environments? There is certainly a debate in the industry about how much intelligence and architecture is needed within a single data-center environment, but there are still network management elements that would allow dynamic optimization of a network for mixed workloads or traffic types.
Also, large clouds will be globally distributed. How will OpenStack support that from a networking perspective? How will VPNs, VRFs, VLANs, MPLS connections, and so on be provisioned and operated in a highly distributed, dynamic environment--if at all? There is a huge opportunity for both service providers and infrastructure vendors to innovate in this space.
And that, ultimately, is why the networking effort in OpenStack will be important to the industry as a whole. Yes, there are other open-source cloud management systems worth considering for your cloud system, and yes OpenStack still has a long way to go before it reaches the maturity of several of those other offerings.
But, for whatever reason, the industry is converging on OpenStack for now, and that portends great things happening in the coming years, especially when it comes to the role of networking in cloud services.