In “Is OpenStack Dead,” we focused on OpenStack’s future as a public cloud platform. We concluded that Amazon, Microsoft, Google, and VMware would dominate public cloud computing through complete economic and technical control of their stacks. We further concluded that OpenStack would have to be free to cloud computing vendors in order for them to have a financially competitive offering, and further, that IBM, HP, Rackspace, and Red Hat do not have an economic incentive to invest in creating a free cloud platform for service providers worldwide.
The Private Cloud
Before we discuss who is going to win and lose in the private cloud market, let’s first examine why any company would want to stand up a cloud of its own when it can go rent one from Amazon, Microsoft, Google, or VMware. Reasons might include:
- Most enterprises have on-premises data center virtualization environments. VMware leads this market by a wide margin with vSphere and the vCloud Suite; it is followed by Microsoft with Hyper-V and Red Hat with KVM.
- Enterprises have transient workloads. Development, test, pilot, and education and training are all workloads that need to be instantiated, run for a period of time, and then torn down. Automating the lifecycle of a transient workload with a private cloud and a service catalog makes a ton of sense for the IT department and for the users of these services.
- IT departments are under tremendous pressure to improve their agility and responsiveness while refraining from growing, or even while shrinking headcount. Automating the provision of a wide array of services allows IT departments to better serve the business while containing costs.
- Offering a robust and easy-to-use private cloud to business constituents can “stop workloads from walking out of the door,” which many IT departments view as an existential threat.
- A private cloud allows IT to offer a standard set of services that transcend pure Infrastructure as a Service (IaaS). Middleware platforms like Cloud Foundry can be offered up as a service. Desktops can be offered up as a service. Entire multi-tier applications can be offered up as a service.
The Hybrid Cloud
A hybrid cloud allows you to seamlessly move workloads back and forth between your own private cloud in your own data center, a cloud that you own but that happens to run in a data center that you rent, and a true public cloud. So, the promise of the hybrid cloud is that, with maximum flexibility and convenience, workloads can be moved around based on where they are in the lifecycle and on what their production execution requirements are. An enterprise might want a hybrid cloud if:
- It has some applications that are completely unsuited for public cloud deployment and some that are a natural for public cloud deployment and wants to run each application in the place best suited for it.
- A new application with security or performance requirements is developed in a public cloud, like Amazon, and the business wants to bring it back in-house to run in production.
- The enterprise wants to build and test a new application that requires elasticity in-house but then to deploy it at Amazon, because Amazon does a great job of running applications that are very elastic in a cost-effective manner.
The Key Attributes of a Successful Private or Hybrid Cloud Platform
If you are looking for a private or hybrid cloud platform, here are the criteria upon which you should base your decision:
- Choose a vendor that has a proven ability both to create a competitive private cloud management offering and to continue to innovate with that offering. This is where VMware and Microsoft have a huge advantage over everyone else, as they already have huge installed bases of on-premises data center virtualization software (the prerequisite for a private cloud). Many third-party vendors also have compelling offerings: among them are Virtustream, which specializes in performance and security guarantees that no one else offers; CloudBolt Software and Embotics, which specialize in cloud management software that can be installed and put in production quickly; and ElasticBox, which focuses on provisioning entire application systems across internal and external clouds. None of the vendors behind OpenStack (IBM, HP, Rackspace, and Red Hat) have a credible track record in the private/hybrid cloud.
- Address the question of compatibility, and therefore seamless workload migration, between your public cloud and your private/hybrid cloud. Again, VMware, Microsoft, and Virtustream have huge advantages here, as they all own their own public clouds and offer identical software (or nearly so) for on-premises deployment. Eucalyptus has a play here as the on-premises version of AWS. If your private cloud and your public cloud are going to be different, look at offerings from companies like HotLink and Racemi that can easily migrate workloads back and forth. OpenStack does not fare well here, as it has neither a large installed base in the public cloud nor a large installed base in the private cloud, and it is incompatible with all other public and hybrid cloud offerings.
- Make sure that your private/hybrid cloud offering can automate the entire set of services you want to automate and deliver. Compute in a service catalog is not enough. Your cloud should address software-defined networking, storage, and application platforms (PaaS) and have the ability to deliver entire applications as a service. OpenStack again fares poorly here, as it currently lags behind the market leaders in cloud management functionality, and it shows little sign of being able to catch up.
- Make sure that your cloud platform is easy to implement and easy to support. This is where OpenStack completely falls down. If you are not willing to devote a substantial number of software engineers to your OpenStack project, it will never get off the ground. OpenStack is a box of Legos. Mature cloud management offerings are easily configurable solutions to a variety of automation problems.
Is There Any Hope for OpenStack in the Private or Hybrid Cloud?
The best way to understand this is to understand the motivations of the vendors behind OpenStack. IBM and HP would very much like to establish private clouds in their enterprise customers based upon their OpenStack distributions. But all of these same enterprise customers now run VMware. And the cost of switching from a very mature VMware offering to an OpenStack offering that requires a far greater headcount to support is something that most enterprises will choose not to endure. Rackspace has no on-premises software business today and is unlikely to build one. Red Hat (with RHEL) has a credible business supporting Linux in enterprise accounts worldwide. If Red Hat can actually productize OpenStack into a cloud management offering that can compete with the commercial offerings from VMware, Microsoft, Virtustream, and Eucalyptus, then Red Hat OpenStack has a chance. But Red Hat has a lot of ground to cover to catch up, and it is unlikely to get a lot of assistance from IBM, HP, or Rackspace in the pursuit of its own success.
A private/hybrid cloud management platform must offer a rich set of services, be easily implementable, and offer compatibility with public clouds. OpenStack fails in all of these dimensions and as such is destined to completely fail as a private or hybrid cloud offering. Therefore, OpenStack will fail as a private, hybrid, and public cloud offering.