The coming years could see the cloud computing market rise to over $100 billion because many organizations have found that the cloud offers great benefits over using in-house hardware. For example, saving money is one of those many potential benefits. The advantage of the cloud is paying for systems and computing resources “as you go”, while also offering a completely scalable environment. Organizations can scale both up when business is booming, or down when fewer systems are needed to manage the workloads.
Different Clouds for Different Situations
The first ever public clouds introduced features that the general IT market and new DevOps consumers found attractive. These services involved spinning up virtual machines complete with selected OS and coupled storage systems and a guaranteed level of service. The common benefit for these first generation clouds included infrastructure agility and compute capacity for when you needed it.
High Performance Computing (HPC) users, on the other hand, have a different set of requirements including tightly coupled compute clusters over a low-latency network. HPC applications often have been painstakingly tuned to run on bare metal systems without the overhead of a hypervisor. In addition, tuned hardware might have even been selected based on processor speed, memory and communication for each application set, with tuned storage, that has been designed for specific applications and setting the user base.
Because organizations have diverse requirements for HPC workflows that have often been highly tuned, it is important to understand the economics of managing in house resources and the options that are available in the public cloud market. The cloud computing market is still young, but growing fast and evolving quickly, which means that you’ll get the best results from the cloud when you do your research and test your ideas. Having a cloud provider that understands the class of service required by HPC applications is important because it offers the most direct comparison of the economics of cloud with in house HPC clusters.
When to Move from In-House to HPC
Many HPC administrators understand the economics of operating large clusters. Budgeting for datacenter space, power, cooling, compute, storage, network and staff have been part of standard business process for decades. However, as the cloud market matures and diversifies, the advent of hyper-scale HPC cloud services brings a new set of tools for organizations who rely on HPC for a competitive advantage.
Historically, most commercial organizations invest in HPC over standard depreciation cycles which can be as short as 3 years and sometimes as long as 5 years. However, cloud service providers are investing in state of the art computing technology at two to three times the rate of most enterprises. When coupled with available scale and purchasing power, the economics can often tip in favor of a “buy” versus “build” strategy. It also does not have to be an all-or-nothing choice. Many organizations have started leveraging hybrid strategies where there is a blend of on premises HPC that is tightly coupled to the public cloud.
Whatever stage you are at in your evaluation of HPC, you must understand your individual use cases, applications and workflows. This will help you to maximize the economic benefits from HPC investments. The cloud offers a higher degree of flexibility than what you might find with in-house servers, and you can also choose a hybrid solution to reap the benefits from both ends of the spectrum. Your organization can take advantage of the many applications offered on the Nimbix Cloud including ANSYS, CST Studio Suite, and CD-Adapco’s Star-CCM+.