Cloud computing inches toward open source
OpenStack is a good start for a project still too complex for DIY
For a technology that's all about openness, cloud computing is pretty proprietary.
Cloud-service providers including Verizon, Terremark and Bluelock run on VMware's cloud software. Microsoft's Azure runs a Windows-like environment using Hyper-V. Others run home-grown cloud platforms that range from Amazon's massively scalable cloud to the barely foggy virtual-server cluster at a local hosting provider.
Rackspace, Cloud.com and the Openstack project have all released cloud platforms with varying levels of openness as a way to let end-user companies and service providers build cloud platforms without relying on major software vendors.
Openstack, a six-year-old open-source community backed by the major cloud provider Rackspace, which also claims Cloud.com as a member, operates with the goal of creating standards to make cloud platforms interoperable.
Ideally that would mean end users would be able to move virtual machines from one cloud to another; more likely in the short term is that it would provide APIs or protocols that would allow apps or virtual machines on more than one cloud platform to be managed from the same console.
Right not it's not even possible to move a virtual machine image from one cloud running VMware software, for example, to another cloud running VMware because the location and configuration of services such as databases and authentication would be different on the new platform.
Openstack has separate modules for compute- and storage-sharing, which essentially install on top of physical servers and do a complex load-balancing dance to allow those servers to contribute CPU cycles or memory or storage to application workloads that may be running on virtual servers on another physical computer in the same data center, or one located somewhere else but connected through a storage or data network.
"What LAMP was to the data center, OpenStack will be to the cloud," said Alan Shimel, managing partner at The CISO Group and a Network World blogger.
The cloud software itself is actually based on the Nebula cloud platform created by NASA and Rackspace to give NASA a more efficient way to use its own resources.
Rackspace backs Openstack with funding and development resources, and uses Openstack software as the basis of its own service.
Rackspace adds full integration and consulting and additional features such as the content-delivery service CloudFiles, training program called CloudBuilders, as well as its own hosting service.
Cloud.com contributes code and development help to OpenStack, and uses OpenStack as part of its CloudStack partially open/partially proprietary infrastructure.
It tries to set itself apart with additional layers such as CloudStack Business Logic – which lets service providers using CloudStack as their platform bill for the privilege.
It also provides connections and support for other clouds using AJAX clients, APIs of its own, and a module called CloudBridge that's designed to integrate CloudStack with Amazon's EC2 using SOAP and REST to translate EC2 API commands into native CloudStack commands.
Cloud.com also supports three major virtualization platforms – Vmware's vSphere, the open-source Xen owned by Citrix, and KVM hypervisors.
The potential to share workloads between clouds, or move workloads from one to another is still at least a year off – more likely two or three. Being able to exchange data and execution requests is a major step, though.
So is the potential to use open-source code to build a private cloud without mortgaging the company to one of the big hypervisor vendors.
A December InfoWorld test of five private-cloud products made it clear building a cloud under any circumstances is no picnic.
Trying to do it, right now at least, using immature open-source software, would be the kind of adventure you probably don't want to bet your career on.
Kevin Fogarty writes about enterprise IT for ITworld. Follow him on Twitter @KevinFogarty.