Many companies that move production apps to the cloud wind up moving them back on-premises again. In this interview with The Enterprisers Project, Laz Vekiarides, CTO of cloud storage provider ClearSky Data, explains why this happens and how to make cloud deployments that will stick.
The Enterprisers Project (TEP): Why do companies that have moved production apps to the public cloud bring them back to their own data centers? What frustrations or obstacles are they hitting?
Vekiarides: In our experience, we see this happening because production apps need to be cloud-native in order to be successful in the public cloud. Often, they need to be written from scratch with the public cloud in mind, and this isn't a standard practice in the industry, yet.
The public cloud is not intended to be a highly available, predictable entity, like a clustered server farm that lives in a company-controlled data center. Performance can be unpredictable, things fail all the time, and some large providers randomly conduct maintenance on servers, which directly affects the users hosting workloads in those locations.
It's a far cry from the five-nines world of on-premises infrastructure. The resulting experience can take companies by surprise, as can issues such as hidden costs associated with rewriting production apps for the cloud. It's easy to see why some companies dive into cloud strategies headfirst, only to realize later that they're not fully prepared to take every step needed to make those strategies successful.
TEP: How can those frustrations be addressed?
Vekiarides: Working with cloud-native production apps will bypass many of the issues associated with public cloud deployments, but not every company has the bandwidth and resources to rewrite or replace existing apps. There are also many products on the market offering targeted solutions that aim to ease one aspect of the cloud migration process, but IT teams should be wary of any single-use solutions that might only complicate infrastructure without addressing holistic problems.
The best option in today's market is to work with managed service providers that write services and solutions from scratch with the cloud in mind. At its core, the IT industry needs a generic system layer that sits above public cloud infrastructure, and guarantees that the cloud is enterprise-ready, with the high performance required by enterprise data centers.
TEP: Tell us a bit about infrastructure-as-a-service (IaaS). What are the advantages? How does it really work?
Vekiarides: To me, IaaS is that system layer that allows enterprises to effectively consume public cloud. The most basic definition of IaaS could be as simple as a traditional MSP deploying its services. In the cloud era, however, IaaS is a technology that helps companies leverage the low-cost, pay-as-you-go capabilities of the public cloud.
In general, IaaS is a major opportunity for service providers because IT pros are attracted to service-based consumption models. IT environments are expanding quickly and becoming more complex, fueled by big data solutions and enterprise cloud adoption trends. Meanwhile, security threats and data center consolidation initiatives are putting pressure on legacy hardware.
Services help companies escape the "rip and replace" IT mindset and that can cut costs and improve overall efficiency. For the same reasons a family might choose a Netflix or Hulu subscription over a traditional cable network, IaaS providers are recognizing a market need for core data center offerings and expert support, delivered on an as-needed basis.
TEP: Cloud deployments, particularly public cloud, seem to be the way of the future, yet the public cloud has not yet fully lived up to its potential for replacing application servers and storage in-house. How do you see cloud usage evolving in the coming years?
Vekiarides: The public cloud is simply not the best answer for every company and every workload. However, hybrid computing models offer an economic value that is too appealing for any enterprise to ignore. In the next few years, we're going to see more serious cloud migration experimentation and creative variations of hybrid strategies. Eventually, we'll likely have a set of public cloud providers offering pools of computing power across global locations, and companies leveraging those resources in new and interesting ways. In general, however, companies will stop acquiring the bulk of their infrastructure as a capital expense, and will make it an operating expense instead. And they'll find new ways to make cloud processes valuable.
