It’s widely acknowledged that the combined ingenuity of the Amazon Web Service team and the creative and fast-moving Xen community led to the creation of the first Infrastructure as a Service (IaaS) Cloud – Amazon EC2 – as far back as 2006. So it has been interesting to see VMware attempt to co-opt the concept and turn a profound capability that was made possible by the open source community into a proprietary enterprise software play. It’s not that I don’t admire VMware for laying claim to the concept of cloud – I do – and to be sure they’ve done a good job selling expensive virtual infrastructure products to the enterprise. But as any fan of General Patton could tell you, if you’re going to claim technical leadership, you need to deliver or get out of the way.
VMware has been playing the “we invented the cloud” drum-roll for a couple of years now, but has yet to deliver its first enterprise-ready cloud product – vCloud – which is being developed in a project codenamed Redwood. It is likely that vCloud will be formally announced at VMWorld later this year, but with an announcement today of the OpenStack project, the open source community has re-asserted its claim to leadership in cloud computing, offering a complete open source cloud infrastructure platform developed by a powerful community of collaborators that share a commitment to free, open source, hypervisor and API-agnostic, standards-based cloud infrastructure that will meet the needs of enterprises and service providers alike.
The OpenStack effort is chaired by Rackspace whose Cloud Servers and Cloud Files implementations will be contributed to the project, and NASA, whose Nebula cloud orchestration layer will form a key component of OpenStack. The Rackspace Cloud is built on XenServer – the Citrix supported version of the open source Xen Cloud Platform, which itself draws on over 20 other open source projects including kernel.org and openvswitch.org. But a key tenet of OpenStack is openness and as such it will be hypervisor neutral, and while Citrix will be a leading contributor to OpenStack, the strength of the OpenStack effort is its large contributor base, and the fact that it contains both leading clouds and infrastructure vendors. The full list of participants at the OpenStack design summit held last week include: AMD, Autonomic Resources, Citrix, Cloud.com, Cloudkick, CloudSwitch, Dell, enStratus, FathomDB, Intel, iomart Group, Limelight, Nicira, NTT DATA, Opscode, Peer 1, Puppet Labs, RightScale, Riptano, Scalr, Sonian, Spiceworks, ThoughtWorks, Zenoss and Zuora.
The OpenStack community is founded on a belief that proprietary, closed source, single-vendor cloud offerings such as vCloud are wholly at odds with the concept of cloud computing, which can only succeed if openness is mandated at every interface. And we are not the only folks who think this way. Tim O’Reilly, the famous tech pundit and founder of O’Reilly Media is quoted on the OpenStack community site thus:
“If cloud computing is the future, then understanding how to make that future open is one of the great technology challenges of our day. Rackspace and NASA are taking an amazing step towards my vision of an open cloud future.”
Today, then, is a day for VMware to recognize the creativity and leadership of the open source software movement, and its role in providing a vehicle to support the shared business goals of more than 25 ISVs that stand ready to deliver on the true promise of cloud computing – something that VMware can never offer. Whereas Redwood promises to offer a modestly scalable cloud orchestration platform with single-vendor support for the VMware vCloud API and VMware vApps and VMs, the OpenStack community will offer a completely open sourced cloud infrastructure, for enterprise or service provider use, with all of the key attributes that customers ought to expect of every cloud infrastructure – whether public or private. OpenStack brings:
- A commitment to compatibility, workload portability and openness: OpenStack will support both de-facto cloud APIs from major providers and emerging standards from the DMTF (including OVF, which VMware calls vApp, and which it has already begun to extend in a proprietary manner), and other standards organizations. Moreover, OpenStack will support a range of API bindings. If you insist on talking to OpenStack using VMware vCloud API, that will work too. And the community will ensure that OpenStack based clouds will federate with cloud resources from any other provider. In addition OpenStack clouds will be hypervisor agnostic, and (courtesy of XenServer and the Xen Cloud Platform, as well as organizations such as CloudSwitch) will run any VM from any hypervisor.
By comparison today “…there’s no simple way to use vCloud with vSphere 4 …VMware is working on that capability but the current vCloud APIs don’t mix and match with existing VMware installations. For instance, vBlock users cannot smoothly shift workloads and virtual servers from vSphere to vCloud Express”.
- Scalability and security born of real world deployments: Whereas project Redwood will probably scale to support a couple of thousand servers, and likely will be useful for private clouds, OpenStack will benefit from the contribution of code from the industry’s largest clouds such as RackSpace’s and those used by the scientific community at NASA, that demand scalability to tens or hundreds of thousands of servers, with guaranteed correctness in the face of complex systems faults and security born of real world deployments in hostile environments. There are now several XenServer based clouds in excess of 10,000 servers in size, and the management architecture for such a system is radically different from that of vSphere and what I’ve read about Redwood – which both suffer from many single points of failure.
- Architected for multi-tenancy: One of the most important requirements of any cloud infrastructure is the need for isolation of different tenant workloads on the infrastructure. In the enterprise, multi-tenancy enables IT to guarantee that different IT user groups or applications are properly isolated, which in turn is a requirement for regulatory compliance including PCI. At the virtual infrastructure level multi-tenancy begins with the hypervisor and its security architecture. Beyond security, the platform needs to offer guaranteed SLAs to different workloads, and a way to account for resource. XenServer (and the open source Xen Cloud Platform – XCP) delivers everything that is required here. Beyond the hypervisor, the virtual network must guarantee isolation, and resource guarantees. In the case of OpenStack, the new open source virtual switch that today is part of XCP and that is developed at openvswitch.org provides unparalleled granularity of control as well as an open architecture for control and management of the virtualized fabric. Finally, multi-tenancy extends into storage, which can take the form of block or object stores. XenServer manages the block store and the Rackspace Cloud Files implementation offers an object store.
- More than enterprise virtual infrastructure at scale: Whereas VMware’s focus with vCloud is to get its virtual infrastructure platform to scale, and to offer APIs for virtualized compute, networking and block storage, OpenStack includes the Cloud Files object store from Rackspace, and in so doing enables any enterprise to implement its own private S3-like object store, with the choice of which objects to host privately and which to replicate into the cloud. OpenStack will offer a fully-featured cloud service suite to enable next generation cloud-based applications to scale elastically, with granular resource control and full SLAs.
- Engineered for cloud, from the get-go: One of the principal benefits of cloud-based infrastructure is the fact that it provides granular information for charge-back, to enable IT administrators or service providers to appropriately bill for services delivered. OpenStack has been engineered to provide powerful role-based management and granular accounting for resource use, from the get go.
- Support for differentiated, value-added services: Cloud service providers that implement vCloud will face two nasty challenges. First, all cloud providers that implement vCloud will be simply offering the same services as their competitors. Second, they will certainly pay a lot more money for the privilege of doing so than they would by adopting a free virtual infrastructure platform such as XenServer or KVM, and the OpenStack suite. OpenStack does not attempt to define the entire cloud, and leaves plenty of room for service providers to add and integrate (courtesy of its openness and open source code base) other service offerings.
Finally, it’s important to note that OpenStack is not the only open sourced cloud stack available. Cloud.com, Abiquo and Eucalyptus all offer open source orchestration platforms today. If OpenStack is to succeed it will need to continue to draw on the contributions of the ISVs that have joined RackSpace and NASA in the announcement. Citrix is certainly committed – more clouds running XenServer, OpenVSwitch and our value-added VPX virtual appliances is good for us – the commoditization of the cloud orchestration business will leave several vendors puzzling over their strategies. I’ve counted over 20 such vendors in the market, with new entrants showing up all the time. Most recently Nimbula, founded by ex AWS execs, joined the scene with respectable investors.
Whilst the ultimate winners in the cloud infrastructure vendor category remain to be chosen, it is clear that the business of cloud orchestration will be commoditized. What’s fascinating is that in this case the key commoditizers are cloud vendors themselves, who recognize the importance of openness and compatibility, and the need for a platform that supports value-added features that enables them to compete. More importantly, for enterprise customers, with the ready availability of entirely free virtual infrastructure and storage platforms, the addition of open source orchestration means that the business of building an open, compatible cloud of your own will be simple, and that there is a clear set of alternatives to VMware’s lock-in agenda.
It’s time for VMware to admit that as far as cloud is concerned, the game is up, and that its monopolistic, ISV-antagonistic business model has no play in the infrastructure as a service market. The open source community brings to the table a richer, more secure, more scalable, and open platform not simply because it has an idealistic notion of cloud, but because it taps into the strengths of a diverse ecosystem of contributors, each of whose business interests are best favored by the rapid adoption of an open cloud architecture. If you’re an enterprise customer of VMware, OpenStack promises to serve you better than vCloud will, and if you’re looking to purchase cloud services, OpenStack and other open source based clouds will offer you a richer set of features, with a proven security model and better performance than you will find from service providers that simply adopt vCloud to offer their customers just another way to consume VMware licensing.