Nmedia - Fotolia
SAN FRANCISCO -- Enterprise cloud computing professionals are marching over a bridge. On the one side is research and evaluation and on the other, cloud deployment that exploits the inherent benefits in the architecture.
"The burden on enterprise IT is to be economical and deliver IT effectively, while also being the business and supporting the business [in a digital economy]," said Correy Voo, the infrastructure CTO at global financial company UBS.
There's a latent demand for the exciting things that only IT can do for a business, Voo said, speaking as the president of the Open Data Center Alliance (ODCA) at the group's Forecast 2014 event here.
The existing volume of IT infrastructure isn't capable of handling this demand, Voo said, so there's an accelerated opportunity for companies in cloud to deliver value.
Private enterprise cloud, or, as Voo defines it, utility computing, presents daunting integration tasks to IT teams because of its complexity, according to ODCA member survey data. Interoperability of legacy equipment is another major hurdle to overcome. Customized deployments go against the whole reasoning for cloudification, Voo said, so the industry needs to share deployment models, reference architectures and other acquired knowledge among peers. Voo spoke with SearchCloudComputing following his Forecast keynote.
Why the shift in focus from evaluation to cloud adoption?
Correy Voo: The Open Data Center Alliance membership drove this change. Over the first three years [since its inception in 2010], the emphasis was on sharing knowledge, documentation and standards. As we go into the next stage, we have to ensure the work and intellectual property being shared around the community is actually used.
We're looking at how many people put enterprise private cloud plans into action. Are people understanding and using that [intellectual property] effectively?
External clouds are designed to address the needs of the masses, not the needs of businesses. The enterprise feature set isn't in the public cloud today, or your regulatory requirements on IT aren't met by public cloud providers today. This doesn't mean large enterprises can't have cloud -- they just can't have public cloud. Large-scale businesses with legacy environments are big on private cloud. And private cloud doesn't necessarily have to mean on-premises cloud anymore; it could be a self-contained isolated infrastructure at a third-party provider.
Private, public and hybrid cloud deployment schemes are beginning to blend. The new hierarchy -- I call it utility computing -- is a cloud developed and built internally, unique to my business. There are users with a customer base -- think of the major retailers or financial institutions -- that create a big enough market to warrant that private type of cloud infrastructure.
What should change when you adopt a cloud deployment model? What shouldn't change?
Voo: Adopting cloud doesn't give you the right to give up all your previous best practices for IT management. You wouldn't give a five-year-old child a $700 iPhone without supervision. You'll hurt yourself if you just let everyone consume cloud resources without guidelines -- they're just different guidelines [than traditional IT deployments].
There's more opportunity for chaos in a cloud deployment that isn't bound by physical limitations like data center size, or the number of people on staff. How do you make scalability and elasticity practical? There are deployment and usage models for these scenarios. Maturity will help with that. Cloud isn't mature in any way yet.
What should enterprises consider if cloud adoption is on their roadmap?
Voo: Start out on base principles. Be very clear on whether or not you need cloud. Not everyone will. What do you want to adopt and why? What will it do better than what is in place now? Cloud computing greatly helps situations with high client-side mobility, massive amounts of data, or a lot of application mobility where you don't know or care where the application lives.
What should stay out of the cloud? Selection criteria are much easier once you know what you expect from cloud infrastructure in terms of agility, control, data permissioning and other results. If you must be deterministic about what you do with the data and who is running it, the applications don't lend themselves to external cloud, but a private/utility cloud deployment can help. And you can build this enterprise cloud to [the] same scale if not greater than what you would buy in an external cloud.
Don't forget about SaaS. Software as a service [SaaS] gives you an opportunity to take a business process and buy into a set of cloud-enabled tech with improvement to processes inherent to cloud deployment. You're swapping the environment rather than migrating. If I have a bad internal application, I'll go buy a SaaS product that does the same thing. It's analogous to moving house rather than remodeling.
The IT industry still aspires to happy shared computing. We have new acronyms and terminology to describe features and concepts that have been around since mainframe's heyday. If I say "flexible resource" -- that's understood. Say "cloud" and the meaning might be more nebulous.
What should every cloud deployment model cover? How can companies share model expertise?
Voo: Find someone else who has seen and done this before. The ODCA created its Cloud Expert Network for this reason. Large-scale multinational corporations through small startups all want to make cloud work and all want to adopt cloud, and have no alternative to meet the goals of IT-enabled business. The prospect of going back to traditional computing is just impossible.
Enterprise cloud deployment always raises concerns about data privacy and management/security. That situation is [exacerbated] by high-profile cases -- think iCloud leaks. You have to bring forth your own plan for cloud security, rather than the CEO taking the lead without built-up technical knowledge. We can't miss out on educating the consumer base for our clouds. The IT team exists so the business can sell things to its customers. If they have concerns and fears because they picked up info from the general media, they're going to transpose those fears onto their providers.
Share learned best practices for the basic mechanics: integration, interoperability, sustainability, stability. We have to do these things as IT guys anyway, but in cloud deployments they become more acute. To get the real benefit out of an enterprise private cloud, you have to think about those things in a different way. It's not a technology question -- it's the softer aspects [such as] people and process change. Technology issues can always be resolved. Usage models focus on operational change because that's where the technology goes right or wrong. Practical guidelines are great for big companies and even better for smaller companies.
There's inertia associated with legacy; nearly every company of any size has to deal with legacy. When you have legacy infrastructure, you have to contend with it while implementing a new world. You may be enthusiastic about deploying cloud, but the maturity worries people.
Final thoughts on cloud deployment today?
Voo: There's a point of reflection where there's so much emphasis and expectation placed on cloud that we owe it to ourselves as consumers and technologists that we set cloud on the right path. The competition that's occurring now is in some respects detrimental. You want the benefit of movement and change. That evolution and flexibility will occur in the next three to five years, where if you want to leave one cloud infrastructure for another, you can without a hassle.
Get ready -- really ready -- for cloud adoption
Build a better enterprise cloud than AWS