The role of cloud computing in today’s enterprises continues to accelerate, fueled by both market pressures to compete more aggressively against digital-savvy competitors and, more recently, by the COVID-19 crisis, which has prompted a massive shift to digital work and consumer engagement. While it’s clear that cloud computing has a vital role to play in supporting new and existing applications, it presents difficult choices for data managers.
There is a case to be made for keeping some data on-premise, moving other datasets to the cloud, and pursuing both on-premise and cloud strategies simultaneously. Adding to the mix are approaches that increasingly involve multiple cloud platforms. Where is all the data going, and where should it go?
ACCELERATING INTO THE CLOUD
The pace of data movement to the cloud continues unabated, especially for new applications and systems that are being rolled out. At the same time, there will be data on-premise for a long time to come, as well, said Grant Fritchey, product advocate and data privacy expert at Redgate Software. “For existing organizations, and even for new ones, the trend leans heavily toward a hybrid approach with some data in the cloud and some still stored in a more traditional manner.”
There is a “data gravity” force that is inexorably pulling things toward the cloud, said Lindsay Marshall, manager of cloud solutions for SAS. “Data gravity is real, and as the volume and complexity of data continues to grow, and as more enterprise applications move to the cloud, so does its pull of data toward the cloud,” she said. “As business requirements change, the elastic nature of the cloud means organizations can quickly and easily scale up or down and only pay for the time used to carry out specific activities.”
The movement toward cloud is seen across many industries, and financial services is no exception. “The number of firms either planning, designing, building, or running workloads has increased dramatically over the last 3 years, as has the pace at which workloads move from proof of concept to production,” said Cory Albert, global head of enterprise data cloud strategy at Bloomberg.
Parts of this movement are fueled by cloud vendors expanding services to many relational and NoSQL databases. “Based on these market trends and the demand that we see in protecting these environments, we believe growth in the adoption of cloud-deployed databases will accelerate,” said Param Kumarasamy, vice president of product management at Commvault. “We’re seeing increased demand for cloud database data protection support as customers like to manage on-premise and cloud workloads on a single pane of glass.”
MAKING DATA CHOICES
Of course, there is no single formula for what stays in the cloud and what stays on-premise. There are situations in which enterprises can control the flow of data to the cloud, and situations in which they may not be in control. When data migration is planned, the choice of the cloud itself may be in an enterprise’s control—as is the case of explicit migration to AWS S3, said Rajesh Raheja, senior vice president of engineering at Boomi, a Dell Technologies business. Conversely, he said, in some cases, cloud choices “are not in their control—such as when subscribing to a SaaS offering that may store its data in any cloud of its choice.” There are even partway choices, such as virtual private cloud instances from public cloud providers or co-location-based data centers on-premise, Raheja noted.
In the case of unplanned, or accidental, data movement, “individual lines of businesses may initiate use of SaaS offerings without adequate governance at the corporate level and end up with many disparate systems of record,” said Raheja. “This is particularly true in vertical areas, such as marketing, where many niche vendors exist to satisfy different functions. If a company uses even a few SaaS vendors, it will lead to proliferation of data across multiple clouds, since each SaaS offering may be storing the data in their own data centers or use one of the public cloud providers,” Raheja said.
MAKING THE CASE FOR ON-PREMISE
That leaves the question of whether some types of data are better left on-premise. The case for cloud may be compelling since “the resiliency, scale, elasticity, provisioning, lifecycle management, management and monitoring maturity, availability and disaster recovery abilities, all point toward keeping data in the cloud,” said Raheja.
However, the risk profile determines where data resides. “Information governance and the adherence to industry regulations greatly influence a company’s willingness to be cloud-first,” said Kumarasamy. “If their risk profile is low, then we see a higher propensity to make a complete migration to the cloud.” However, if the company’s risk profile is high, it will be reluctant to move data to the cloud. The degree of data protection offered may help shape this decision, he added.
At the same time, cloud providers may offer greater security than corporate data centers are able to provide themselves. “Vendors providing cloud-based solutions must now comply with stringent data security regulations from external third parties, ensuring that risk is effectively mitigated,” said Marshall. “Security in the cloud is all about covering the bases, continually improving and benchmarking against current best practices, and encompassing people, process, and technology.” Still, in some situations in which organizations must comply with regulatory requirements, they may choose to leave data on-premise, Marshall said. She cautioned, however, that security is but one dimension of this question. An enterprise may have “legacy applications with interdependencies with other applications and databases, and applications dependent on traditional or monolithic database structures,” she said.
The decision about keeping at least some data on-premise, then, may require taking “a holistic view of the data and applications in your current IT environment, and tiering them in order of importance and criticality to operations,” said Sam Roguine, director at Arcserve. For starters, he explained, the speed at which it is necessary to recover these systems and operations will highly influence whether data and applications get stored on-premise or in the cloud. “If a server fails, for example, it can be too costly and take too long to recover from the cloud due to network latency.”
Cloud-based storage, Roguine continued, makes sense for “recovering from wide-area issues, like natural disasters, that can knock out entire data centers. It differs on a company-by-company basis.” Hybrid cloud arrangements allow for freer transfer of data and applications between on-premise and the cloud, giving IT managers free rein over which datasets live where, he explained.
Data best left on-premise “is that data which simply won’t move well to the cloud,” said Fritchey. “Think about large-scale data systems that require robust, dedicated hardware to manage.” These may be highly transactional in nature or very analytics-focused, but the data is better managed locally than in the cloud, he said. “One example is a hedge fund that needed to capture and analyze data faster than is possible through cloud systems. They tried, but their in-house team just does a better job. However, even there, some of the data that is less mission-critical has moved to the cloud, creating a hybrid solution.”