For all the attention it's been getting, OpenStack is still in its formative years and the future success of the project lies not in how it will change future IT operations but how it will mesh with existing IT.
Developing a list of enterprise cloud requirements can be difficult. My first experience discussing cloud management platforms with enterprise software vendors was an enlightening experience. I had a fixed idea of private cloud. My concept included a self-service portal and the ability to move existing workloads to a VM farm with a cloud portal for measuring usage. I basically wanted VMware vSphere with (the now defunct) VMware Lab Manager. After installing an early release of OpenStack I quickly realized that my idea of cloud didn't coincide with the OpenStack community’s vision of cloud.
After installing OpenStack and not getting a portal to provision VMs, I had to dig deep into what OpenStack provided the enterprise infrastructure. What I soon learned is that OpenStack wasn’t about the infrastructure operator or VMs, but the application developer. OpenStack provided a set of AWS-like APIs to the infrastructure. Development organizations looking to build Web 2.0 applications look toward cloud management platforms that provide robust APIs to developers.
There’s a disconnect between what OpenStack offers and the requirements of the infrastructure group that holds the keys to implementation within the enterprise. I don’t believe my original concept of a private cloud is very different than many VM-focused infrastructure groups. Today’s OpenStack can meet these requirements but may be overkill based on the effort needed to install and maintain the platform. VMware vRealize would be a better fit for a basic use case, such as a self-service catalog.
No shortage of OpenStack investments
What’s certain is that enterprise IT vendors have been making major investments in OpenStack and have made significant acquisitions of existing OpenStack vendors. IBM purchased managed OpenStack provider Blue Box and Cisco purchased Piston Cloud Computing, maker of a popular OpenStack distribution. While some have labeled this a coming of age of OpenStack, I’m not as convinced. I view it as needed consolidation in an unfocused open source project.
I view the diversity as one of the challenges with the OpenStack community. Since OpenStack isn’t a product managed by a single entity, there’s a lack of focus. Consolidation helps eliminate the number of voices and desperate agendas. EMC’s VP Technology Randy Bias provided an excellent State-of-OpenStack presentation at the last OpenStack Summit held May 18 in Vancouver. One of the more interesting points Bias makes is his desire to see more project competition. Bias not only called for increased competition but for the foundation to allow projects to fail when they are not meeting the requirements of the community. For example, he noted the notorious reputation of Neutron for not meeting the network automation needs of most users. I’m hopeful that Cisco’s acquisition of Piston Cloud results in increased participation and a much improved Neutron project or replacement.
What's driving OpenStack and what needs to change
During HP Discover, fellow bloggers and I had a Q&A with representatives from HP’s Helion product and technical teams. Helion is HP’s custom distribution of OpenStack. The majority of projects that went from pilot to implementation were driven by new application deployments. For me, this is an indication that IT infrastructure teams are driving OpenStack deployments for general consumption by their developer customers. What are the requirements that OpenStack is missing for wider adoption by enterprise IT?
I think it goes without saying; OpenStack is still very hard to implement. EMC's Randy Bias made this an abundantly clear point during his presentation. While no private cloud is easy, OpenStack is particularly difficult. Specifically, OpenStack calls for skill in areas that most infrastructure teams aren’t adept – the most obvious being development. OpenStack developers need to have an understanding of infrastructure technologies along with development skill. The combination is not easy to find in individual developers. It’s not a light statement when OpenStack proponents say it only takes about a dozen developers to support PayPal’s private cloud. I agree with Bias's assessment that the industry needs to make the OpenStack deployment less reliable on customization via developers.
Infrastructure groups aren’t ready to provide AWS-like APIs to the entire data center. One of the primary deterrents is that the existing enterprise data centers aren't designed to provide cloud services. LUN based storage arrays, AIX hosted applications and non-programmable networks are all examples of barriers to the private cloud. Traditional data center services such as storage, Unix and network are not API-friendly. Provisioning storage, for example, is an extremely manual process. A human is needed to understand where applications should be placed and how LUNs should be laid out based on application requirements. Processes that require human intelligence are nearly impossible to automate and therefore impossible to consume as a cloud-based service.
Companies such as EMC have been focused on providing an in-between platform while OpenStack is a green field approach to the private cloud. EMC’s Federation Hybrid Cloud is an example of this approach. EMC supports both the traditional VMware-focused infrastructure while providing an API for building next generation cloud applications. In our LUN-based example, the EMC model would provision all storage to VMware vSphere. At this point the storage is virtualized. With all storage virtualized, the cloud manager can provision VMs based on API requests. The EMC model doesn’t prevent end users from managing a portion of the infrastructure using legacy processes for legacy applications. A successful OpenStack enterprise approach has to take the existing data center into consideration.
While the EMC acquisition of Randy Bias’ CloudScaling and IBM and Cisco acquisitions don’t validate OpenStack, the acquisitions provide hope that these traditional enterprise companies can bring their understanding of enterprise IT to a maturing OpenStack.