Those of you familiar with Nimbix know that the Dallas-based startup is challenging convention – wildly with the Nimbix Lounge Party during SC15 this year, and widely. In fact the company is a Texas star(t-up) on the rise – attracting industry watchers, brand-name customers, partners, and new investors. Driving their ‘Supercomputing made superhuman’ strategy and delivery of the company’s revolutionary high performance computing cloud is Nimbix CTO, Leo Reiter whose influence as an industry change agent is extensive. A pioneer in open source virtualization technologies, Reiter’s career has been steeped in startups for years, including the launch of his first company Virtual Bridges, which was acquired by Nimboxx.
Reiter will present a featured Founder Story at the StartupHPC Summit, Monday Nov 16 at the Capital Factory in Austin. We sat down with him to get his perspectives on the challenges, risks, and rewards of startups in HPC.
insideHPC: As a startup/self-starter, business leader – what inspires you to invest your career/focus/company/talent in HPC?
Leo Reiter: For me personally, I’ve been interested in “big compute” my entire life. When I was a kid I used to dream about owning a supercomputer company, mainly so I could design them. I was never that amazed by end user computing or any sort of personal computing, not even mobile technology. I’ve always been drawn to the types of problems that can be solved with tremendous scale and speed – in other words, what HPC is all about.
From a pure business perspective, it’s “the right place at the right time”. It’s clear that the challenges we face on a global scale are simply beyond the means of end user and personal computing. We have to process enormous amounts of data, and do it quickly, in order to fight disease, find energy sources, and design better, bigger products and structures. The intersection of Big Data with HPC, what IDC called “High Performance Data Analysis”, opens things up even more.
HPC is no longer for traditional technical computing problems, but rather, to tackle the greatest challenges humanity has ever faced, head on, on a scale we could not previously even imagine. I can’t think of focusing my energy anywhere else.
insideHPC: How do you define HPC?
Leo Reiter: HPC is really about “data transformation”, which is an application of the much-hyped “digital transformation”. We are collecting a massive amount of data. We have moved way beyond just searching it for information, and into the phase of deriving insight from it. Without the high performance processing, the information quickly grows stale and loses value. So HPC today is about both big data *and* big compute. We’ll always have the traditional parallel problems to solve, but the real potential is in applying these complex algorithms to the massive amounts of unstructured data produced every moment. HPC will help transform this data into valuable insight.
From a more technical perspective, HPC is about workflows, rather than “work loads”. That means algorithms designed to produce results, rather than services designed around “request/response” architectures. Some HPC software designs are parallel, some are distributed, but all have the same thing in common: scalable architectures leveraging large amounts of compute power to achieve results.
The integration of specialized hardware (such as accelerators and coprocessors), with highly optimized software, transforming data at massive scale – this is how I define HPC.
insideHPC: List your top picks in HPC invention/innovation?
Leo Reiter: Mine, easy …
- Cloud/on-demand delivery of HPC
- Use of accelerators to boost algorithms, such as GPUs and FPGAs
- Convergence of big data with traditional HPC
insideHPC: What do mandates like the recent NSCI mean for the HPC industry/community?
Leo Reiter: Mandates recognize the importance of HPC for the future of society. But we must understand that true innovation will come from putting HPC into the hands of more people, not just making more HPC available to the traditional HPC user.
insideHPC: Why do you think there aren’t more HPC-inspired startups?
Leo Reiter: Three reasons come to mind:
- First. Popular culture makes mobile and other consumer technologies “sexy” – not HPC.
- Second. Many investors have been badly burned by HPC in the past, and avoid funding new startups.
- Third. Access to HPC resources for these startups is either rare or oversubscribed.
The very public and visible success of the web, social media, and mobile attracts more entrepreneurs than to HPC.
Thankfully, many things are changing that should turn the tide.
First, Hollywood is starting to “fall in love” with supercomputing (again) – blockbuster movies such as “The Avengers” and most recently “The Martian” feature prominent roles for supercomputers. Even if these roles are sensationalized, not long ago the stars were mobile devices and personal computers, which are simply not that interesting anymore. Even the portable “communicators” with “universal translators” of past SciFi movies and shows feature technologies that are already years old and ubiquitous. HPC is starting to become “en vogue” again.
Second, one of the major growth areas we’re seeing in HPC is democratizing it, generally via cloud computing; this will put HPC into more hands than ever before, and unleash innovation at the scale that PCs did versus mainframes. This will take away one of the major obstacles to HPC startups, which is gaining access to the systems or technologies upon which to build their wares.
insideHPC: Why are programs like StartupHPC important?
Leo Reiter: I was incredibly impressed with the inaugural StartupHPC – it was simply the best one-day conference I’ve ever attended. The variety of material, the diversity of the speakers, and the engagement of the audience was very impressive. It really was both about startups and about HPC, and was a great way for young, aspiring tech entrepreneurs to get exposure as well as learn.
insideHPC: How do we encourage more HPC inspired startups?
Leo Reiter: We first need to define HPC around problems rather than technology and features. If we get people thinking about curing disease rather than building exascale architectures, we’ll attract more entrepreneurs.
Investors should start believing in the convergence of big data and big compute, and getting excited about the new market possibilities of HPC.
We need to be much bolder – we need to imagine a future not bound by current architectures, and then figure out how to solve those problems – instead of first understanding/accepting the limits of our technology, like we do today! And we need to stop talking about exascale – this is something most people cannot even comprehend.
Let’s focus on solving more complex problems faster, and if that demands exascale, we’ll get there (see first recommendation).