One can be forgiven for thinking that everything that is anything IT-related is being absorbed into the cloud. There is immense pressure to make the cloud move. As found in a recent survey of 550 executives conducted by IDG, more than one-third (38%) report their IT departments are under pressure to migrate all applications and infrastructure to the cloud.
However, it's notable that at this point, most corporate IT and data assets still sit in their own data centers. There are some applications and systems that simply aren't a good fit for cloud at this time -- and may never make it into the cloud. How do IT decision-makers determine when a cloud migration doesn't make sense, and it's better to leave things on-premises? Previous iterations of the IDG survey had found 56 percent of executives intend to stay with their on-premises systems for the foreseeable future, and at least 40 percent say they are too dependent on mission-critical legacy systems to migrate to cloud.
To explore the issues of when staying on-premises versus cloud makes sense, I asked industry executives about any areas that were not suitable for cloud, and better left on-premises -- especially from the all-important data perspective. The security implications, as well as geographical presence requirements, are obvious. But there are also other facts that may make staying on-premises the most viable option. Here are some of the concerns raised:
Legacy entanglements: For starters, there may be a very complicated entanglement of on-premises assets that making moving a particular piece to the cloud a complicated project, says Hal Woods, chief technology officer at Datera. "If a given application interoperates heavily with other applications that must run on-premises, then that application and its database must also run on-premises, naturally," he advises.
Abay Radhakrishnan, CTO architect for Sungard Availability Services, agrees, cautioning that "applications that integrate with the database and are tuned with particular runtime or server for specific database versions" are better left on-premises. Plus, he has also seen examples of on-premises applications that "are tuned with specific versions of their databases on certain platforms for performance and reliability reasons."
Monolithic legacy applications can also simply be a bear to migrate. "Legacy applications and applications with monolithic database structures are still best suited for on premises," says Scott Harvey, vice president of engineering and operations for Atmosera. "Generally, legacy applications are not compatible with the cloud -- especially with some of the abstractions found in public cloud, such as SaaS or PaaS offerings removing some of the more system admin-focused management and configurations, or, in the case of some PaaS offerings, not offering parity with on-premises installations of database engines."
Cloud sticker shock: Escalating cloud costs is another case for keeping certain assets on premises. Datera's Woods relates that "some public cloud providers are known for ingesting data at a reasonable cost, but they charge a lot for egress. After getting a whopping cloud provider bill, many organizations look toward taking some of their databases back on-premises, as was the case with one of our customers."
Security: Data itself can be complicated, and this makes cloud moves untenable, says Dave Nielsen, head of ecosystem programs for Redis Labs. "Some data can be exposed publicly, and some data cannot risk even a hint of exposure. In this second case, an air-gapped, on-premises system may provide better security."
Need for speed: There may also be a need for speed, Nielsen adds. "Large collections of data can take hours, even days, to move from one data center to the cloud. Applications working with large data on-premises should be located near the data where it can better interact with the data. This is especially relevant to modern applications that rely on real-time user data interaction, high-speed analytics, personalization, or recommendation." This may also extend to internet of things configurations as well. "If you are collecting data at a high rate from local IoT devices, going directly to a public cloud may be too slow, so it makes more sense to use an on-premises database or edge cloud," Woods says.
Still, there are companies going all-cloud -- especially those that don't have existing inventories of IT systems already painstakingly built and configured to their particular situations. "We're seeing this mostly at startup companies and in the innovative divisions of enterprises that are working with big data, streaming data, data analytics applications, and new trending technologies such as artificial intelligence, machine learning and blockchain," says Radhakrishnan. "The enterprises that lean more towards cloud use are typically those within industries that have less stringent regulations, such as entertainment, media, travel, or logistics."
While Atmosera's Harvey still sees the largest tier-one applications that require high performance databases being run on-premises, cloud is ideal for "less-demanding applications." Moving to the cloud "has the advantage being faster for an organization to increase and decrease the amount of hardware to scale instantly, versus having to order and install hardware on-premises," he points out. "While past investments in on-premises hardware should typically be utilized, other factors including the administration burden for the infrastructure may entice an organization to move to a public cloud sooner than later."