OpenStack: Time To Capitalize On The Momentum

In the IaaS market the open source torch has officially been passed from Eucalyptus to OpenStack, a community effort that is showing strong momentum in both vendor participation and end user interest. But now it needs to start showing staying power, and that's just what I expect to see at this week's OpenStack Design Summit in Boston. What started as an effort to leverage the open community to help advance the technologies started by Rackspace and NASA has now turned into a vibrant community advancing IaaS technologies at a rapid pace. What it was lacking up until this summer was solid go-to-market momentum. But now:

  • Citrix CloudStack (the former has several high-profile wins under its belt including Korea Telecom, GoDaddy and Zynga.
  • Rackspace Cloud Builders, a professional services group acquired by the founding hoster, has set up OpenStack clouds for a variety of service providers and enterprises including MercadoLibre and CERN
  • NASA alums have spawned two new startups planning to distribute OpenStack-based cloud solutions. Nebula, founded by former NASA Ames CTO Chris Kemp, looks to bring OpenStack to enterprise clients in the form of a hardware appliance installed top of rack. Two former NASA Nebula cloud architects last week unveiled their company, Piston Cloud Computing, and will offer private clouds as a hardened OS to enterprise clients.
  • Dell and HP both have announced plans to offer OpenStack-based cloud solutions.

Even more vendors are expected to jump into the OpenStack distribution business this week and we will hopefully learn about even more enterprise and service provider implementations. This is strong adoption for a technology that really didn't have a complete compute component until earlier this year. Thursday of last week (Sept. 29th) the OpenStack community announced the fourth release of OpenStack, code-named Diablo, that brought some much-needed features to its compute element (Nova) and increased the maturity of its other key elements. 

But there is still much to be built out in OpenStack and some key issues under heated debate. Identity management (Keystone) and a consistent yet modularly extensible user interface dashboard are blessed projects for the next release. An ambitious virtual networking project is in official incubation at this stage leveraging works such as OpenFlow, open vSwitch and newer ideas brought forth by Nicira, which could push OpenStack ahead of vendor efforts which are currently pushing the cutting edge furthest - VMware and AWS mostly notably. I'd like to see workload manifests (beyond OVF), and core resource tracking and chargeback foundations added to the portfolio so these base elements of cloud computing can be standardized. 

There are key contentions to be debated at the conference as well. Front and center is the direction of the OpenStack APIs. One camp is pushing to keep the APIs independent in hopes that they grow into yet another de facto standard, distinct from VMware's vCloud and the AWS APIs. They argue for independence as this will give OpenStack more autonomy in feature creation, optimization and differentiation; holding to compatibility with vendor-led closed APIs will hinder innovation. The other camp is fighting for compatibility with these APIs, arguing that doing so will make it easier for management vendors to deliver consistent solutions and invest their R&D dollars in new capabilities rather than mapping functions to each target platform. It will also make it easier for enterprise customers to move workloads between clouds and build hybrid solutions. Both sides have their merits and drawbacks. 

Infrastructure & operations professionals are encouraged to weigh in on these debates, as OpenStack suffers from the same challenges most open source communities do in their early days - lack of enough enterprise voice. If you have an opinion on what you want from private clouds and how they should be designed and implemented  now you have the chance to be heard. 

I'll be leading a panel discussion of enterprise OpenStack adopters discussing these and other issues Thursday, October 6th at 3:30 pm EST. Come lend your voice to the discussion by either attending the design summit Monday through Wednesday, the OpenStack conference Thursday or Friday or by participating in the community at You have the chance in front of you today to steer and shape the private cloud market. Sitting idly by is opportunity lost. 


the deal breaker in cloud solutions

While our wish lists for OpenStack (or for any other cloud solution for that matter) are quite similar, there is still one thing that eludes me and my clients: How do we engineer our private clouds in a manner that can assure / guarantee availability as well as (especially) performance SLAs for mission critical applications?
Dev/Test environments, "slim" web-based applications and non-critical applications alone don't quite cut the ROI numbers for us. Can we really (and how?) deploy performance sensitive applications?

Mission critical too high a bar for cloud today

The key is in setting more realistic expectations for a still-maturing technology. Cloud architectures provide a consistent environment and that level of consistency is still maturing. It's unrealistic, and frankly would be cost prohibitive, to build out a consistent cloud environment at the mission critical level. As for the ROI, start smaller so the return comes faster on a smaller volume of workloads at the non-critical or business critical level. Once you can operate this effectively, then you can start looking at how to do higher levels of HA on the platform and ensuring performance for specific workloads. The cloud solutions coming to market are all working on the performance SLA problem.