1. News
August 11, 2021updated 31 Jan 2022 1:47pm

Supercomputers helped us understand Covid-19. Now HPC-as-a-service is here to stay

By Eric Johansson

Covid-19 changed supercomputing. High-performance computers (HPC) are essentially clusters of connected computers with an accumulated processing power that put today’s fastest home desktops to shame.

These powerful inventions have been around since the 1960s. Over the years, researchers and corporates have used HPCs to develop everything from new medicines to 3D modelling. But during the pandemic things changed. Now they set out to save the world.

Since the start of the coronavirus crisis, companies like Google, Advanced Micro Devices (AMD) and Microsoft have donated and lent researchers support and access to HPCs, enabling them accelerate their research and find ways to fight the disease.

Over the past 18 months, supercomputing has, among other things, helped researchers understand how travel restrictions can halt the spread of the contagion, how virus particles spread in rooms and accelerated the discovery of new therapies against SARS-CoV-2.

In March, the White House launched the COVID-19 High Performance Computing Consortium. The project aimed to bring together researchers and the HPC industry. Today it has over 43 consortium members that have helped 100 projects by donating 6.8 million cores and 600 petaflops of processing power.

One petaflop is 1,000 teraflops. To put that into perspective: the 2019 Mac Pro’s top speed is 5.6 teraflops. These projects arguably greatly accelerated the fight against the pandemic.

“All these modelling and all the experiments that they ran on super supercomputers could have been run on less powerful machines, but it would’ve taken a lot longer,” Filipe Oliveira, senior analysts at GlobalData and the author of a recent GlobalData report on HPCs, tells Verdict.

These efforts also highlight the growing demand of a new type of cloud service referred to as HPC-as-a-Service where customers would be able access supercomputing capabilities without having to buy or build one.

HPC against Covid-19

The COVID-19 High Performance Computing Consortium notably brought together several industry rivals – such as Google, Amazon Web Services (AWS), Dell and Microsoft among others.

“Covid-19 really was a very unique situation and it was a time, [and it’s] still ongoing, where we saw everybody wanting to contribute,” Joe Corkery, healthcare and life sciences director of product at Google Cloud, tells Verdict.

Several of these companies also launched their own independent funds to support researchers during the pandemic.

For instance, AMD launched the AMD COVID-19 HPC Fund in April 2020.

“We had a responsibility as a multinational company to join the fight against Covid, but not just throwing money at it,” Travis Campbell, senior manager of AMD’s external research office and the person leading the charge with the HPC fund, tells Verdict.

While donating money to PPE, AMD also donated “countless hours of engineering time” and 12 petaflops of data over the course of the pandemic. So far it has given 24 projects access to these HPC resources.

The donations and initiatives often took the form of providers giving researchers access to HPC processing powers via the cloud, by installing hardware at the recipients’ institutions or with a mixture of both.

But why did they do that? Businesses exist to make money. Why would they give away processing power and expertise for free? There are a number of reasons, chief among them is the fact that the health crisis hit very close to home. They could see the effects.

Corporate responsibility

Nirav Mehta, director of product management of cloud enterprise at Google Cloud, vividly remembers the moment when he realised how empty the Googleplex in Mountain View, California, was becoming.

“I’m used to the Google buildings having a very strong culture of people having lunch together, long lines and people meeting at the snack centre, and those started to thin out and we started to ask ourselves what’s happening,” he tells Verdict.

Gradually, the buildings became emptier, the infection numbers rose and Mehta realised that something had to be done. This was getting serious.

“I had a pivotal moment where [I realised that] this is serious, we [must] stop and take stock,” he says.

Other representatives from the HPC providers Verdict has spoken with recall friends who’ve suffered from Covid-19 or have seen their livelihoods crumble due to social distancing rules introduced to fight the coronavirus.

To some extent, helping companies this way through the crisis was already part of the services HPC businesses provided before Covid-19. They were set up to help clients meet their potential.

“Google has a strong focus on the user and on doing good, and we wanted to make sure that we were able to leverage the skills, scope and breadth of Google to get us all out of this situation as fast as possible,” Corkery says

Of course, HPC providers benefited from their donation in more ways than just adding to their karma accounts. While they provided their services to researchers, they also benefited from the feedback to their services. In other ways, they learned how they could improve themselves.

“It’s a very mutual beneficial relationship,” Campbell says.

Moreover, it also helped providers demonstrate their capabilities, which will prove beneficial as the supercomputer sector is getting extremely crowded.

Growing competition

Supercomputing is a growing business sector. The adoption of the cloud is part of the reason why HPCs have grown in popularity in recent years and why several providers were able to donate their services in the way that they did during the pandemic.

Thanks to the cloud, businesses no longer need to invest in the infrastructure required to build a local data centre to access HPC resource. Oliveira believes this will lead to a wider adoption of HPC-as-a-Service or Supercomputing-as-a-Service.

“It means that instead of buying all the equipment and investing on your own local data centre to run high-performance computing, what companies do is they buy an HPC subscription service, usually from one of the big cloud providers,” Oliveira says.

In other words, while HPCs have long been the domain of companies that can afford the cost of building clusters of supercomputers of their own, the cloud has enabled businesses to tap into the awesome processing power from these computer behemoths with ease and on the cheap.

MAPRE, Spain’s largest insurance company, has for instance said that tapping into AWS’ supercomputing service was 90% cheaper than building its own HPC infrastructure.

No wonder then that the world’s largest cloud service providers, such as AWS, Google Cloud and Microsoft Azure have recently made moves to capture shares of this lucrative market.

For instance, the Top500 June list of the world’s fastest supercomputers saw Azure grab four supercomputers in the top 30. Comparatively, AWS had one entry on the list, in 41st place.

Using cloud HPC solutions also provide more flexibility. For instance, Richard Lawrence, the Met Office IT Fellow for supercomputing, recently noted that the reason why the Met Office uses cloud supercomputers was because it gave it more flexibility to pick the best solution in future, as highlighted by ZDNet.

Apart from medical and scientific research – such as the donations provided by HPC companies during Covid-19 – supercomputing applications also include oil and gas exploration, computer-aimed engineering, weather forecasting, bringing back supersonic passenger jets like the Concorde, designing electric vehicles, modelling for silicon chips, analyse financial risk and climate change modelling.

Climate change is, incidentally, one of the areas where AMD might refocus its COVID-19 Fund towards as the chaos of the health crisis subsides.

“Covid was obviously, hopefully, a once-in-a-generation or multiple-generations event,” he says. “But it doesn’t mean that that those same unique resources that we talked about earlier can’t help in other really globally complex and challenges right [such as] climate change.”

Intimately linked with the emergence of HPC-as-a-Service is the growing prevalence of artificial intelligence (AI).

“The number of work AI workloads that can run on a supercomputer is a lot higher than what would be possible with a normal computer,” Oliveira says.

He believes that the combination of supercomputers and AI will mean that businesses will be encouraged to develop even better HPC services.

Given how supercomputers helped us get through the black swan event of Covid-19, improving HPCs can only be a good thing to help us fight off future crises.