It has become cliché to talk about “developers as the new kingmakers,” a point popularised by the analyst firm Redmonk. But the point of gifting power (through infrastructure) to developers was never really about developers. It was about changing how enterprises build their products.
And, ultimately, about who builds those products. Yes, developers are central, but as can be surmised from a recent Series C fund raise by Rescale, with Microsoft, Nvidia, and Samsung all contributing, it’s a good time to inspect what others are building on the platforms developers have built. Specifically, engineers and scientists are among the big beneficiaries, and are dramatically accelerating scientific evolution as a result.
Fuelling PhD productivity
But first, back to those developers. With every business scrambling to become a software company, developers are the bright, shiny objects that every enterprise treasures.
No longer cost centres that need to be outsourced, development teams are now seen as top-line revenue generators powering the business. Leadership has learned to invest in developers as growth drivers. The rise of OPEX cloud computing has fuelled this trend.
But developers were never the end goal, per se. Developers were simply paving the way so that others could contribute more fully to enterprise productivity. Specifically, it’s time for engineers and scientists to grab some of that developer spotlight. These highly paid specialists (typically PhDs) have grown accustomed to having to queue to run massive workloads on bespoke on-premises hardware, which costs a fortune and slows the pace of inquiry. Then it’s back in line to iterate what they learned from the last job.
That was then, this is now.
Increasingly these same engineers and scientists can turn to the cloud and accelerate the iterations of their workloads that simulate real-world conditions to get more innovative products to market much faster. It’s the principle first laid out for me years ago by Matt Wood:
Those that go out and buy expensive infrastructure find that the problem scope and domain shift really quickly. By the time they get around to answering the original question, the business has moved on. You need an environment that is flexible and allows you to quickly respond to changing big data requirements. Your resource mix is continually evolving; if you buy infrastructure it’s almost immediately irrelevant to your business because it’s frozen in time. It’s solving a problem you may not have or care about any more.
Hard problems — like autonomous vehicles, rockets, and supersonic transport — benefit from engineers and scientists being able to flexibly mould infrastructure to the questions they’re hoping to answer.
Boiled down, smart companies have learned that the best way to attract and nurture developer talent is not only to compensate them well, but also, and more important, to remove obstacles in their work.
The rise of SaaS (with an API for whatever back-end function you need), Jamstack, Kubernetes, and all these other new technologies spreading across the enterprise software stack free developers to focus on the logic of the new application or service they are developing. They can forget about the infrastructure. Time-to-market cycles speed up. More and better services delivered much faster leads to happier, stickier customers. And more top-line revenue.
In sum, it’s a partnership between developers and engineers/scientists. Developers abstract away all the infrastructure hassles and suddenly your engineers and scientists can help your business beat the competition and grab market share. It’s a match made in heaven. Or Hacker News.
Distributing cloud benefits to the PhD set
Back to Rescale and its investors. Microsoft (venture arm M12), Nvidia, and Samsung are all long on cloud. Microsoft wants more workloads on Azure — bonus points for higher-margin HPC jobs — and Nvidia and Samsung want to sell many more higher-margin, specialised chips.
The genius of Rescale, and other new startups in this HPC cloud-brokering marketplace, is that by aggregating workloads across their big HPC customers they can achieve the scale that makes it possible for AWS, Microsoft Azure, and Google Cloud to accelerate the CAPEX investment required to support more complex, specialised HPC/AI/ML workloads (specialised chips, ultra-fast I/O, etc.).
This creates a flywheel effect. It pours gasoline on long dormant industries like automotive and aerospace (largely consolidated decades ago) and helps ignite innovation.
That’s a bold claim, given just how much these three cloud vendors spent on CAPEX in 2020 alone. By Charles Fitzgerald’s estimate, they collectively spent $97 billion on CAPEX last year. That’s not all for their cloud businesses, of course, but a considerable chunk of it serves that market. As Fitzgerald says, this money is “bonkers” big.
But size only matters inasmuch as it conveys real benefits on other industries, which, it turns out, it does.
Suddenly it’s easier to understand why private industry startups can hurl rockets into space after mere years of development and prototyping (hundreds of space startups were funded in the past five years alone) when it took decades to create Saturn V or the Space Shuttle at orders of magnitude more expense. In aerospace today, the United States’ primary strategic bomber capability, the B52, flies with an airframe that was designed by engineers with slide rulers back in 1948.
The priests in the HPC temple want in on this action now. You can’t do hard digital research and development (rockets, genomics, smart cities, etc.) without smart scientists and engineers. These very smart people watched their developer friends in the enterprise free themselves from the shackles of infrastructure. Now they want their own intelligent software-defined computing connected to fellow researchers, scientists, and engineers for collaboration. Sounds like a cloud use case to me.
HPC cloud brokers like Rescale sit in a unique vantage point for these scientists. They know what job you ran, what software you use, what version under what licensing terms, what interconnect, what data you put in, and how long the job took.
They also know exactly how much it cost — and can tell you how much it will cost for all your other jobs. At the same time, they can arbitrage the best price-performance for your workload across their network of partner cloud providers.
They solve a very hard problem for customers solving the world’s hardest problems. Customers can also ask, “What is the best software for the job I want to run?” That is a compelling value proposition for algorithmically-driven workloads and widely diverse use cases.
Best of all? No more PhDs waiting in line to run their jobs.
In the world where engineers and scientists live today — solving mind-boggling challenges of extreme complexity — they are starving for change. They want to solve problems, not understand the complex infrastructure that makes a legacy data centre look like a child’s Lego house by comparison. Rescale and its peers promise to give the PhDs full-stack HPC economics/performance optimisation, and deliver on that promise continuously.
In sum, there are things that developers increasingly take for granted, like continuous integration and continuous delivery (CI/CD). For HPC engineers and scientists, however, such things are relatively new — and somewhat miraculous. In the hands of the PhDs, they’ll undoubtedly lead to miraculous things.