In his recent blog post, Toby Owen hinted at future technology to extend our cloud into customers’ data centers. We’re working toward that end, but in the near term we want to allow our customers to create their own networks in our public cloud. To accomplish this we’re adopting cutting edge technology around Software Defined Networks (SDN). I’ll discuss our implementation of SDN more in future blogs, but for those of you at Interop this week in New York, Toby and I will present an overview of how we are using OpenFlow-based logical switches in our hypervisors. (Our session, “Taking Cloud to the Final Frontier – SDN and Network Virtualization – Presented by Rackspace Hosting,” is Thursday, October 4 at 11:00 a.m. eastern.) Managed by a Nicira controller, these switches allow users to create logical networks between compute instances.
We have given select customers and an army of Rackers access to these networks, developing new use cases for the public cloud. One Racker used three networks and six VMs to make a decent sized BGP network with Quagga as a cheap network test lab. Others are using broadcast and multicast on these networks to experiment with High Availability (HA) clusters. Some Rackers and network appliance vendors are also experimenting with VPN and firewall virtual appliances as compute instances. The versatility of these environments has been astounding, even though a limited feature set is currently accessible via API.
I’ll save more details for the future, when I can speak freely about how we are going to productize this network capability. For now, come to our session at Interop on Thursday and see the future of networking.