The evolution of high performance computing

Penguin Computing
Add to MyDirectIndustry favorites

As a Linux infrastructure provider specialising in high performance computing (HPC) technology, Penguin Computing works with many different types of engineering firms to help them get the computing resources they need to model their designs and produce revenue, writes Director of Cloud Services Will Cottay.

We’ve seen a distinct pattern emerge in the last twelve months of firms focusing on increasing efficiency to better support long-term growth. Different sized firms as well as those that focus on a particular type of design have some specific HPC but even their technical choices are manifesting themselves in some fairly specific trends.

Rethinking the cloud

Project-based engineering work is inherently unpredictable, especially at firms working on more than one project at a time. However, investing in the powerful HPC resources needed to do modelling and simulation is a major investment. For some types of firms, it can even be a show showstopper.

Small to medium firms that focus on manufacturing (such as a firm that produces automotive armrests) have historically been at a tremendous disadvantage, starting from when they want to bid on a contract. Execution of most contracts require significant amounts of data (on what an armrest might look like out of the mould, what happens in a crash, etc.) but you can’t get that data without doing simulations (since repeatedly creating and destroying designs is too slow and potentially too costly).

However, small to medium manufacturing firms may already be using their HPC resources for existing projects or cashflow simply won’t allow them to buy more resources ahead of a contract they may not even get. This dilemma is so severe that it inherently limits growth so engineers have tried a variety of options to get around it, none of which really resolved the issue. Fortunately, more non-virtualised HPC-on-demand solutions have appeared and been proven to be powerful enough to meet engineering needs. As a result, more firms are willing to rethink the cloud as a way to solve these problems.

We see a similar trend among engineering departments inside larger companies that manufacture highly regulated or otherwise complex products, such as drug delivery systems (e.g. inhalers and syringes) that require verification of design tests. Even large manufacturing engineering firms with huge HPC needs are rethinking cloud services simply so they don’t have to get into the HPC space themselves.

In both cases, they may have access to on-premise HPC resources but find the return on investment (ROI) of a hybrid solution more appealing than in the past. They are excited by the idea of transparency and predictable costs as well as the freedom to focus on design thanks to these powerful non-virtualised HPC cloud platforms. The ROI is now potentially so good that firms that already have hybrid cloud are looking at the ROI of moving fully to the cloud.

Moving to remote visualisation

Another trend we’re seeing in modern engineering firms relates to how they interact with data.

Virtually all engineers want to be close to their data so they don’t have to move it back and forth from their computing resources to their workstation and lose time. However, this requires a significant investment in workstations. For small departments in large engineering firms, having many workstations to tap into at any given time is usually both a need (because of the complex, high-touch projects clients often bring to large firms) and beyond the budget.

At the same time, the issue of how to understand and modify models is complicated by modern training techniques. Engineers generally have limited experience running HPC because of its relatively recent widespread usage. However, younger engineers today are typically trained on GUI-based workstations that their well-funded universities had access to - but that many engineering firms do not.

That’s why firms with an eye to the long-term are opting to invest in GUI-based remote visualisations solutions rather than traditional workstations. Using a cloud-based, virtual workstation removes the need for firms to buy, maintain, and physically store workstations. At the same time, instead of young engineers having to learn command line interface-level skills before they can be productive, there is literally an ‘easy button’ engineers can push on a workstation that runs the simulation modifications the engineer has just specified.

On-premise challenges for large firms

Very large engineering firms have a unique mix of challenges that they are trying to address. Many large firms prefer an on-premise HPC solution for security and to support a steady, high demand for resources.

However, those with engineers in geographically remote locations struggle with how to manage user access. In the past, each office basically acted as a unique HPC centre, but that meant HPC resources were not being used to their fullest by the firm overall. To resolve that, we’ve found that some of these firms are building internal HPC clouds.

The desire to keep data secure and on-premise has also led to inefficiency in data visualisation. Specifically, the number of projects in a large firm, and thus the number of engineers, is significant. However, buying, maintaining and physically accommodating so many workstations that are not in continuous use is efficient. Fortunately, the same internal cloud capability that these firms are deploying for access to HPC resources can help their geographically distant engineers visualise data remotely.

We’re also seeing some large firms go beyond an on-premise cloud. It’s still on the horizon, in development, but we see indications that some very large engineering firms are ready to go to automated workflows in their on-premise cloud. By this I mean that they have not only their private cloud but they are looking to standardise HPC usage in that cloud with consistent processes mapped to their organisation’s needs in a way that reduces inefficiency and optimises results.

This new maturation in thinking about how to use HPC technologies is really exciting for the sheer quantity of options that it opens up for engineering firms. We’re really looking forward to seeing how these trends play out and what comes next for the world’s engineering firms.

The evolution of high performance computing