Whilst I have been away on vacation, something fairly interesting has happened in the area of Open Source initiatives for Infrastructure as a Service in the form of a new initiative from NASA and Rackspace called OpenStack. You may remember in our last post in this area, we noted that there was a proliferation of offerings in the IaaS space, and it was in the customer’s best interest for there to be effective migrateability (or even mix and match) amongst public and/or private clouds. However, the API standards to support interoperability are proving elusive.
We identified a major reason for this as being the difficulty of driving cloud standards from a software perspective. By this we mean that the main actors are the IaaS service providers rather than their software suppliers (be they commercial licensors or open source communities.
The next phase of standards may thus be an Open Source “definition by implementation”. A platform like Eucalyptus could conceivably dominate through ubiquity in the same way as the Apache web server. However, Eucalyptus is following the MySQL-style Dual License approach, generating a revenue stream from the proprietary licensing of the Enterprise Version, and using a non-permissive open Source License (i.e. GPL) for the core thereby effectively disadvantaging third-party monetization. In addition, Eucalyptus alone controls the destiny of its product. There is no independent governance model, like there is with Apache, Mozilla or Eclipse. If you want to contribute to Eucalyptus, you must assign IPR (Intellectual Property Rights) to Eucalyptus, and then are granted a license back. At the end of the day, if Eucalyptus goes bust, there’s a problem. You have the source code, but it is essentially like getting it out of escrow. The practicalities of continuing development are daunting.
NASA had been using Eucalyptus as the basis for its own IaaS infrastructure known as Nebula. It claims that it wasn’t able to get it to scale to the required level (you may or may not choose to believe the macho talk of 60 Million VMs) and more pertinently it couldn’t effectively commit and control the destiny of enhancements it was contributing to the open source. As a result it dumped Eucalyptus and built its own IaaS provisioning layer called Nova, and this along with Rackspace’s CloudFS scalable persistence layer (analogous to Amazon S3) forms OpenStack.
OpenStack has decided to go a radically-open route, and thus in our view is a much more serious play for the ubiquitous IaaS platform. It has clearly been put together by people who know what they are doing. It has
- An open API (obviously).
- A single open codebase (no “enterprise” versions).
- A permissive license (Apache 2.0) to ensure a level playing field amongst commercializers
- A contribution model that does not require assignment (essentially it is just a signed Apache 2.0 License)
- Multiple intial core contributors (Rackspace and NASA)
- A major US government-funded initiative (NASA) to help ensure longevity of support (Aeronautics programmes naturally have extremely long support lifetimes).
- Open development (via Launchpad)
- A governance model – actually this is currently embryonic. It looks like Rackspace is providing the legal entity with which the project resides, but there is talk of an open foundation.
- An ecosystem initiative – they’re actually hiring for someone to manage this…
Nova in particular is Beta software, so don’t expect it to work seamlessly out of the box, but there is enough resource behind it to get it polished in time for ubiquitous adoption as the IaaS market expands and matures.
The Ecosystem contains a number of competing IaaS providers including Rackspace (obviously), Peer 1 hosting and Cloud.com (who scurried in as soon as it was announced), and it has EMEA and APAC coverage from Iomart (UK) and NTT (Japan). From a Virtualization Practice perspective we see a number of cloud tools and technology providers such as Rightscale, Zenoss, and Enstratus (who are all betting on adding value to the infrastructure once it is ubiquitous), and we also see Citrix, as a major player in the initiative.
The question remains what is the motivation of the prime movers in this project?
As far as NASA is concerned, is IaaS really rocket science? However, it’s actually NASA’s job, as a side effect of the space program, to develop technology (like the Teflon on frying pans and the “Memory Foam” on my new bed) for the benefit of the world.
Why are Citrix there? Well, there is no doubt that Citrix will be planning to add value to this infrastructure through products. They will most-likely be looking at the way that Microsoft technologies can be deployed into it, and they are also promoting the use of XenServer as a hypervisor (rather than KVM). They are also probably going to be interested in how this plays out with VMware, who are noticeably absent, and given Citrix’s whole XenServer initiative is now positioned as a torpedo into vSphere, no doubt OpenStack will be positioned as a torpedo into vCloud, and there is a real opportunity to sink that ship.
The most interesting participant is RackSpace. Rackspace claims to be number 2 in the IaaS marketplace, we assume this is behind Amazon. Rackspace’s revenues are close to a billion dollars a year, although a lot of this is still in pure collocation where it adds little value. IaaS computing is much higher margin, and a strategic goal to grow the company profitably. Rackspace differentiates itself by the quality of its support and actually doesn’t care what the API standard is as long as there is one. Hence it will be dumping most of its own API in transitioning to OpenStack.
We noted in our last post that hosting companies typically have very high levels of customer loyalty (Rackspace’s churn is 0.8% compound per month – or less than 10% per year), and to the extent that standards would encourage customers to change suppliers they are unlikely to promote standardization, and yet RackSpace has decided to go this way. It may well be that it is seeking prime mover status in the open source initiative, as part of a standards play to overtake Amazon, which will be seen as a proprietary locked-in platform, as will vSphere-based solutions such as Terremark.
We don’t expect to see Amazon playing in the same Open Source project, but they may respond. We also await Eucalyptus’s response, and the possible involvement of IBM. There is also another similar existing initiative known as opennebula, which has not gained that much momentum, but might be used as an alternative vehicle.