Existing, highly integrated applications and systems probably should stay on premise, Nair continued. “Legacy applications that have a lot of integration with other in-house systems and have a lot of data transfer needs between various systems in different locations, are best suited for on-premises database systems,” he said. “Moving such an application to the cloud would require re-architecting the application, ultimately resulting in high costs due to network egress in the cloud.”
Most observers agree that it’s not prudent to attempt to move an entire enterprise environment into the cloud. “There definitely are enterprises putting their database systems in the cloud,” said Dheeraj Remella, chief technologist at VoltDB. However, he added, “I would consider putting all the database systems in the public cloud a bit naive or outright irresponsible. While the public cloud vendors have gotten their data centers certified for the various certification requirements like PCI and HIPAA compliance, it still poses a risk that is not fully measurable as an on-prem data center with hardened borders.”
In many other cases, the move to the cloud has been deliberately slow, as many legacy systems need to be uprooted and migrated. A hybrid approach offers a bridge between the two worlds, related Richard Bradbury, vice president of strategy and alliances at Quore. He noted that his company works with large enterprises in the hotel industry, which is moving to the cloud in a big way. “But with hundreds of systems, vendors, hotel owners, and operators involved, that move is slow,” he explained. “Most have chosen to do it organically—one system group at a time. Each move requires months to years of analysis. But the large chains see the long-term benefits of central management, lower cost of on-site servers and maintenance, more consistent consumer products through monitored and enforced standards, and competitive advantage, or parity, when attracting new franchisees.”
In other industries, hybrid environments may serve to delineate between real-time and latent services. “Our IT organization utilizes a combination of technologies and solutions to manage our internal data,” said Narayan Venkat, vice president, data center systems at Western Digital. The company retains a large quantity of data that is generated from its design, development, and manufacturing teams which it uses to gain insights about its products and processes, Venkat noted. “The majority of our data is stored on premise, but we leverage the cloud for infrastructure and applications to analyze massive datasets that are 18- to 24-months old. Our big data analytics platform is run in the cloud while our business-critical, real-time applications are hosted on premise. This architecture saves our company millions of dollars.”
The divide in a hybrid environment may occur between systems of record and systems of engagement. Rugg said systems of record are often best maintained on-premise, while systems of engagement may be the best candidates for cloud. “In the cloud, you have more flexibility to scale up and down on demand, but you tend to have less control over achieving the absolute best performance or availability in the sense of traditional failover mechanisms,” said Rugg. “For this reason, systems of record tend to benefit from being deployed on-premise. Also, since these systems may have been in place for a long time and are very stable, there is reluctance to introduce the risk of migrating them. These also tend to have predictable workloads and therefore see fewer benefits from the flexibility of the cloud.”
Systems of engagement, on the other hand, that “tend to be web-facing or mobile, and where you have an unpredictable number of users, can benefit from being hosted in a cloud where you can scale out the environment on-demand,” Rugg said.
Hybrid cloud introduces complexities as enterprises seek to keep data in both cloud and on-premise environments, as required by privacy mandates or considerations. “Many enterprises seem to gravitate toward retaining control of regulated data and then turning to off-premise clouds for data or capabilities that do not contain or require personal data,” said Jeff Cherrington, VP of product management at ASG Technologies. “A payment transaction processing enterprise, for example, may keep the core payer demographic and detailed transaction data within the firewall, while integrating the capabilities of a public cloud business intelligence SaaS for analysis and presentation. In these instances, the underlying databases can even be disparate.”