APPS MADE POSSIBLE
What applications or capabilities today would not be possible without cloud services? What is too hard to do with on-premises systems? “Data management, analytics, and AI applications go hand in hand with cloud services,” said Ghai. “Capabilities include real-time data visibility and data quality management.”
Cloud enables data managers to greatly expand their data resources. “The data at the enterprise level is too big, too complex, and too expensive for on-prem,” Ghai added. “The ability to efficiently process and understand large data quantities is key for success and growth.”
Cloud services also “provide robust tools and frameworks crucial for developing machine learning and AI,” said Bruce. “This allows organizations to build, train, and deploy machine learning models without substantial infrastructure investments.” Flynn said cloud answers many of the challenges that beset data architectures of the past. “In all my days, there’s one sight I’ve never witnessed—a hardware procurement cycle that dances to the rhythm of need,” said Flynn. “Imagine a cycle that dynamically senses demand, procures storage, procures compute, delivers it, configures it, tackles workloads, and then scales back down when not needed.”
EXCEPTIONS TO THE CLOUD RULE
Not everything should be “cloud-first,” of course. As industry leaders point out, there are instances in which data and applications should remain on-prem. It’s not advisable to pursue a “cloud-only path,” said Flynn. “It’s vital to explore all options, weigh the risks, and figure out when it truly makes sense to dive into the cloud. Flexibility and a well-rounded strategy should always be on the menu.”
AI-ready enterprises “should be looking to ‘cloud-also’ strategies versus ‘cloud-first’ strategies,” said Sherbak. “The classic flexibility and agility that made cloud computing so popular can now often be matched by in-house, on-premises solutions. Cost benefits of the cloud are also not automatic. Enterprises need to evaluate their existing IT environments to explore a combination of cloud and on-prem options that make the most sense for supporting their core business goals.”
There are many instances in which enterprises should be cautious toward cloud-first data policies. “While they have many benefits, it also depends on your enterprise’s specific use cases and security goals,” said Regensburger. “For example, having a cloud-first strategy for data has a lot of appeal for enterprises looking to get value from their data faster than you would with an `internal, home-grown solution. That said, with a cloud-first approach, you lose the ability to control data the way you would if it was all within your internal data environment.”
When it comes to setting architectural strategy, “there are no absolutes,” O’Dell agreed. “Despite the evolution and adoption of cloud capabilities, the reality is there will still be a need for on-premises data solutions and safeguards, especially in the public sector. While cloud-based platforms provide benefits such as scalability, flexibility, and cost efficiency, many agencies within the public sector, along with regulated industries like healthcare with highly sensitive personal information, may still require on-premises solutions.”
Many organizations are opting “for a hybrid approach that allows them to leverage the benefits of cloud capabilities while maintaining essential on-premises protocols,” O’Dell continued. “This approach allows organizations to strike a balance between innovation and control.”
Data privacy and cloud security are also ever-present concerns that require a great deal of attention. “Business leaders find themselves under relentless pressure from shareholders to safeguard company and customer information, making the shared responsibility security model offered by cloud players less enticing,” Flynn said. “Running operations in a secure, compliant manner within the cloud will cost a pretty penny, and it’s not without its share of risks.”
While most data-intensive functions can be managed in the cloud, “data privacy trumps data gravity and costs,” said Ghai. “When fine-tuning models that require highly sensitive data that cannot leave internal organization boundaries due to legal, regulatory, or compliance reasons, on-prem GPU servers are preferred.”
On-premises solutions may be preferable “in situations involving highly sensitive data governed by strict regulations or when substantial investments have already been made in existing infrastructure,” said Bruce. “But expect that contention to diminish over the next few years.”
Looking beyond the Internet of Things and edge computing, “most workloads can find a home both on-premises or in the cloud,” said Flynn. “However, external factors, such as the ever-evolving privacy regulations exemplified by GDPR, will exert influence on how enterprises safeguard personally identifiable information and make choices regarding their infrastructure."