If you were to read any number of headlines on cloud adoption, you may be tempted to think that the enterprise world is all-in on the cloud. However, you'd be wrong.
Sure, some 73 percent of organizations have at least one application or a part of their infrastructure in the cloud, according to an IDG report. But that's only a part of their application portfolio or infrastructure: many companies still maintain some sort of on-premises infrastructure.
SEE: Research: Cloud vs. data center adoption rates, usage, and migration plans (Tech Pro Research)
In the age of the cloud, what's the point of an on-premises data center? Despite the hype, the cloud is very much still a nascent technology. While cloud security and performance, in many cases, can match that of on-premises technology, there are some instances when it's better for companies not to migrate their apps and infrastructure to the cloud.
"All workloads are different, and each has its own set of business requirements and benefits," said David Cappuccio, a distinguished vice president and analyst at Gartner. "Without taking these factors into account when evaluating an application's or service's eventual placement, negative business impacts would far outweigh the benefits of any migration if the wrong choice is made."
When considering a move to the cloud, companies must examine the limitations around compliance, security, latency, recovery time, availability, data location, and other factors, Cappuccio said. After considering those aspects, on-premises infrastructure still makes sense for many companies.
Here are the top 10 things companies are keeping in their own data centers.
1. Predictable workloads
One of the biggest selling points of the cloud is its elasticity and flexible nature. However, that flexibility often comes at a high price, and may not be the best fit for predictable workloads.
"Anything super predictable and consistent will oftentimes be cheaper on-premise than in the cloud," said Brent Collins, global director of the data center practice for World Wide Technology (WWT). "If you can buy equipment and have it highly utilized in a predictable fashion over a number of years, it's probably cheaper to do it on-premises."
2. Legacy homebrewed applications
Any company that has been in business for while probably has some leftover legacy apps that aren't a good fit for the cloud. These types of apps usually require tight security, said Victor Kruse, a solutions architect at PCM, Inc. "Often they're not portable and [are] difficult to refactor," Kruse added. "They didn't evolve with modern technology."
3. Master copies of archived data
With the amount of data most companies are working with, IT leaders must be aware of egress charges -- the cost they'll pay to pull their data out of the cloud. This cost plays an important factor in what data ends up being stored where.
"Because of egress charges many people only keep [disaster recovery] copies in the cloud, and keep their master archive content under their direct control," said Tom Coughlin, IEEE Fellow and president of Coughlin Associates.
4. Niche software
Daniel Barchi, CIO of NewYork-Presbyterian Hospital, said that a "number of applications in healthcare are written for non-x86 operating systems (such as AIX), which is not supported by cloud providers." As such, even though security is handled by end-to-end encryption, this data resides on-premises.
5. Regulatory data
Geo-specific regulations like GDPR often dictate where certain data (especially that of customers) can be physically stored, which necessitates the use of on-premises data centers. "Laws in many countries restrict what data can be stored by companies about their citizens, outside of the country," Coughlin said. "These laws may be met by keeping the data locally in the country where it originated."
There are also specific regulations around personally identifiable information (PII), regardless of its physical location. This impacts healthcare tremendously, Barchi said. Some device manufacturers aren't even incorporating modern designs and architectures into their products, "making it more reasonable and likely for healthcare organizations to store this information on-site," Barchi said.
Enterprise resource planning (ERP) systems are often some of the most complex and mission-critical in an organization. If these are built as giant monolithic applications with heavy data use and intense transaction requirements, they often won't perform well in the cloud. "Nor do they have built-in resiliency, which is needed in the cloud," said WWT's Collins. Also, the associated heavy data services will cost a lot, Collins added.
Kruse also mentioned high-IOPS ERP situations that have a lot of in-memory scenarios. These need low latency because of how crucial they are to many departments, he said. The fear of losing internet connectivity also leads many organizations to keep these applications on-premises.
7. File shares
While the cloud seems to be the perfect place for file shares, some of the clients PCM, Inc.'s Kruse works with found performance issues with this setup. "What happens is they're just dumping it in the cloud without regard for the proper storage and connectivity, which impacts performance," Kruse said. Instead, these are the kind of group drives that should just be on Sharepoint or a similar service, he added.
Despite its nascence, artificial intelligence (AI) is a prime candidate for an on-premises data center thanks to the high IO and low latency requirements needed to render graphics and environments in real time, said Collins. Instead of moving AI workloads to the cloud, "organizations may be best served by colocating the storage and then processing things in the cloud. It also can be expensive to continually try to upload data into the cloud for processing once for ongoing use cases," Collins said.
9. Active Directory services
Even though Microsoft offers Azure Active Directory for hybrid and cloud-only deployments, many companies opt to keep their Active Directory services on-premises, Kruse said, which stumps him. "There's no reason why they can't put that in the cloud today," he added.
10. Disaster recovery for cloud apps
While this may seem counterintuitive, some companies are backing up cloud applications on-premises as part of their disaster recovery plans. "It's a bit backwards, but we're seeing it more and more," Collins said.
- Cloud v. data center decision (ZDNet special report) | Download the report as a PDF (TechRepublic)
- The data center is dead: Here's what comes next (ZDNet)
- Software defined data centers: A cheat sheet (TechRepublic)
- PCIe fabrics in the data center, but Ethernet strikes back (ZDNet)
- Here's why Microsoft is building a massive data center underwater (TechRepublic)