Cloud computing has become ubiquitous, particularly in large corporate environments. Yet, a vast majority of the applications operating in cloud environments — things such as email, customer relationship management (CRM) or other back-office applications — have a relatively low workload demand.

Many companies have assumed that high-performance computing (HPC) applications with high workload demands do not belong in a cloud environment. Today, fewer than 10 percent of all HPC applications run in the cloud. This hesitancy is understandable, as HPC applications are far more complex than email, require a much higher computing capacity and are often the lifeblood of the company.

Things, however, are changing. The number of HPC applications running in cloud environments is growing rapidly. One of the most important developments: COVID-19 diagnostic research using HPC resources in the cloud is underway.

Some estimate the annual growth rate for HPC in the cloud to be as high as 25 percent. What’s changing? In a nutshell, everything. Cloud environments are evolving to more effectively accommodate higher workload applications, HPC applications are evolving to run more efficiently in a cloud environment, and users are more comfortable with the concept of running HPC applications in the cloud.

Companies are also realizing greater economic value from moving some of these compute-intensive applications to the cloud.

For example, we were working with an auto manufacturer that suddenly found itself having to handle a part recall and, in turn, needed a lot of extra capacity for a short period of time. Using its existing HPC cluster to create the extra parts would have had a negative impact on its current production schedule. The company instead moved its short-term recall production demand to the cloud with no disruption to its existing production schedule.

If you’re thinking about moving HPC applications to the cloud, there are three primary options to choose from:

  1. Augment current on-premises capacity with a shared private cloud. This is a popular implementation, particularly as more companies consolidate their IT infrastructure. One of the primary advantages is the enhanced security and control inherent in a private-versus-public cloud implementation. And, companies can choose whether this is an on-premises or a colocated infrastructure.
  2. Augment current on-premises capacity with a public cloud implementation. While this may provide less control, companies can realize the benefits of the newest hardware and monitoring options as cloud providers will upgrade their systems on a far more aggressive timetable than most users can afford.
  3. Go 100 percent into the public cloud. This is an increasingly acceptable option for two of the reasons mentioned earlier: Cloud environments are evolving to more effectively accommodate higher-workload applications and HPC applications are evolving to run more efficiently in a cloud environment.

It’s likely that the ideal scenario for your company falls somewhere within these options. The key is to focus on the up-front research necessary to understand the different computing models, their potential computing capacity, and the economic impact of each on your company’s needs — as well as the business-cycle requirements that the HPC workloads are enabling.

HPC users and the usage models are evolving. In the past, running HPC applications on anything other than dedicated iron was unthinkable. Today, although some use cases will necessitate specialized infrastructures, a flexible computing model has advantages that cannot be ignored and, if done correctly, can provide benefits both from cost and convenience perspectives.