Every cloud computing vendor I meet makes a point of telling me how long it takes the average enterprise these days to provision a new server (for the record, they tell me anything from 2 weeks to 3 months). Last week I met a vendor that made precisely the same point, but its solution isn't yet another compute cloud. CohesiveFT, whose Elastic Server On-Demand service launched in public beta last week, assembles virtual machines to order within minutes, and deploys them to the cloud of your choice — be it Amazon EC2 or your own implementation of VMWare, XenSource or Parallels (see screenshot, after the jump).
The support for data center (and desktop) virtualization platforms is a sign that CFT (full name: Cohesive Flexible Technologies) doesn't believe that the world is about to consolidate its computing onto a small handful of giant compute clouds. Certainly larger enterprises are much more comfortable deploying to their own data centers (or to colocation facilities) at present, using Amazon (if they do at all) as purely a short-term tactical deployment option. The CFT team sees cloud platforms proliferating, with others besides Amazon emerging — and it's ready to support any of them, if that's what customers want. "We are virtualization agnostic," co-founder Alexis Richardson told me. "[We're] just a giant packaging system in the sky."
CFT's 'elastic server on demand' concept envisages production-line assembly of virtual machines, manufactured to order. "Elastic servers are custom application stacks, built from components, virtualization-ready, that you download to test, or deploy to a cloud," says the company. Another tagline is "Z2V — zero footprint provisioning to virtual servers in minutes."
The insight governing this is that the stack deployed at most enterprises today is multi-sourced. However much enterprises want to single-source, the technology proliferates faster than vendors can consolidate. This ongoing complexity is driven not only by continuing innovation, especially from the open source community, but also stems from the need to tie servers into an increasingly componentized infrastructure of monitoring, security, integration and other services. Although most enterprises use only a small proportion of the total available component matrix to build their application stacks, each individual enterprise of course uses a different set. This means there is no single stack or component that's going to become dominant. So CFT is focussing instead on allowing customers to take a pick-and-mix approach to assembling their stack components.
"The challenge of production virtualization is to bring this together in a manageable way," Richardson told me. That's where CFT has concentrated its development and engineering efforts.
- The first step is to select one of CFT's ready-made component repositories, or assemble your own, mixing commercial software, open source or your own IP. By defining its own repository, an enterprise can manage what's made available to its people.
- Next, you choose your components and click to build the virtual 'elastic servers'. CFT describes its rules-based infrastructure for doing this as "a manufacturing plant for building tried and tested application stacks on demand." The automated process takes minutes to complete, and injects services that are used for integration, management and configuration changes after deployment. Then it makes the completed servers available either for download or for deployment to online and virtual computing infrastructures.
- Using CFT's management system, you can then track the deployed component assemblies throughout their lifecycle, logging all configuration changes, and with the ability to reprovision them to another platform at will.
CFT has picked a thin horizontal slice of the emerging cloud computing infrastructure, but it believes it'll be a pivotal slice, one that enables it to become no less than the leading manufacturer of the virtual machines that it sees taking over the world's data centers and desktops.
With a large community of users, CFT will be able to harvest their collective wisdom, extending its rules database as customers and partners extend into new combinations of components, gradually building up best practice benchmarks for server configurations that it can then recommend to customers.
Interestingly, the business model is based on owning the best data because of scale rather than owning a proprietary slice of the cloud. Thus, for example, CFT is releasing information about VcubeV, a technology it has developed based on OpenVPN to connect virtual servers running on multiple hosts within a single VPN:
"In short, it's a technique based on OpenVPN and our add-on that allows us to connect multiple servers (both physical and virtual) located in various datacenters and hosted at different providers into a single address space. It allows applications to run unmodified as if they were all running on hosts behind a single switch. It works even when hosts are behind NAT and/or very restrictive firewalls. Our add-on called cube-routed is a dynamic routing daemon that allows us to have 2 OpenVPN servers for load balancing and failover purposes."
CFT's founders have backgrounds in the financial services sector and are well connected into the open source community, particularly early partners such as WSO2, SpringSource and MuleSource. Based in Chicago, with outposts in London and Palo Alto, it's already got an impressive roster of large enterprises testing its offering and is well worth keeping an eye on. It's one of the most interesting developments I've come across in the cloud computing space — and considering how much is going on with cloud concepts at the moment that's saying something.