There has been a great deal of buzz around the cloud and what it can do for organisations. Increasingly, however, organisations are becoming aware that merely migrating to the cloud isn’t a panacea – and that there are consequences for getting it wrong. Costs can quickly soar, and unintended costs can have a serious impact on the bottom line. Equally, concerns about data sovereignty and accessibility are growing – with cloud performance management issues again highlighted during the Netflix outage of December 2021. How does an organisation know when the cloud is a good option for their HPC needs?

The cloud is a major technological advancement and presents many opportunities for businesses and organisations. CrunchYard pioneered cloud-based, pay-per-use HPC over 11 years ago – and continues to provide services, not only on our own cloud but on public cloud platforms, such as Microsoft Azure.

However, any new technology and advancement can fall victim to its own success.  Often hype, even well-intended hype, can run away with itself, leaving customers with unrealized expectations and unwelcome surprises.

It can be difficult to know when cloud is useful and when it isn't.

It is very easy to feel overwhelmed at the options and finding unbiased information isn't easy.

This is particularly true in the case of HPC due to its compute, storage, and network requirements.

Many organizations have found that "lift and shift to the cloud" come with increased operational complexity and operating costs.  Some have begun to "lift, shift, and drift back to on-premises."

Other considerations include IT management (managing an HPC system typically requires highly specialised (and expensive) IT staff, and managing HPC in the cloud requires specialised HPC staff that additionally require cloud-based expertise); data security and integrity (for sensitive and proprietary data or data subject to sovereignty requirements); and access to the system.

When it comes to cost, we’ve run some numbers, to provide a simple but realistic guide to understanding the economics of Cloud vs. On-Premises HPC. A summarised calculation below shows the cumulative cost of owning an Office HPC system vs. cumulative costs in the cloud.

This is a 256-core system with a login node and shared storage.

The method holds true for larger and smaller systems with similar outcomes.

 Calculations include:

• IT admin costs
• Electricity costs (for Office HPC)
• Cost of capital is NOT included
• Cloud data storage costs is NOT included

This table shows that there is an optimal point for cloud usage vs. Office HPC.

If your system is being used 100% of the time it shows that owning an Office HPC system is more beneficial from year 2 onwards.

In most cases, cloud only makes sense for HPC systems being used for 20% of the time or less.

A clever scenario used by many companies is to utilise a local HPC system to handle the vast majority of the workloads and burst to the cloud for any overflow simulations.

CrunchYard offers HPC solutions that are tailored to your unique requirements and is fit for purpose. Remotely managed, it can include both cloud and on-premises components.

Our Office HPC range of products, built on the robust and adaptable SUSE Linux Enterprise operating system, can be used in any office environment and for virtually any kind of complex computational need. Office HPC is modular and can be expanded and scaled over time – and you can also burst to your cloud resources for temporary or ad hoc busy periods.

In the cloud or on-premises … CrunchYard can assist your organisation with optimised HPC solutions that adapt and scale as you need.