Why HPC Matters

Laurent Herviou, General Manager, HPC & AI Business Unit, Asia Pacific, Hewlett Packard Enterprise

Laurent Herviou, General Manager, HPC & AI Business Unit, Asia Pacific, Hewlett Packard Enterprise

The IT World is a fascinating industry that has enabled humanity to move from a farming and industrial base to a knowledge-base world where everything computes.

Over the past 70 years, most technological advancement happened through the development and application of new technologies, be it hardware or software. Moore’s law has become a benchmark to go further and faster until recently where the traditional CPU architecture reached physical limitations. And here we are, welcome to an accelerated world. Thanks to GPUs, FPGAS and the like, Deep Learning and Artificial Intelligence (AI) have become, almost overnight, a palpable reality; so let’s look closer at what is driving that turn in the industry.

We are living through a digital revolution marking a fundamental shift in how we compute but most of all how we generate and consume data. We are seeing a convergence between the IoT, Edge Computing, DL/AI as well as big data analytics and at the core of it all is HPC (High Performance Computing) or technical computing as I prefer to call it.

We are living in a connected world with more than 8 billion connected devices, which will grow to 20 billion by 2020 if not more, and all these devices are expected to generate in excess of 40 Zettabytes of data. That is more than 3.7 trillion gigabytes or 58 billion 64GB mobile phones and these are conservative estimates. Today everything computes, from your phone to your fridge, to your watch and your clothes and through all the sensors that will become the heartbeats of a smart planet. From high definition scanners and MRIs to next generation DNA sequencers and high throughput satellites or autonomous vehicles, we are soon to live in a data centric world. The data tsunami has just started and it will become our new challenge as data is to become the new IP.

For example, autonomous vehicles will generate 4TB of data per hour and are expected to require 300 TFLOPS of processing capacity to handle the data in real time.

We are living through a digital revolution marking a fundamental shift in how we compute but most of all how we generate and consume data

This will be like having an embedded supercomputer in your car and doing some serious HPC processing at the edge. Now imagine how much data will need to be processed in a central facility for AI training purposes when you have to correlate and cross-analyze data sets from 10 or 100 million autonomous vehicles. At the other end of the spectrum, take the SKA (Square Kilometre Array) project which will become the largest scientific instrument ever deployed with 100,000 antennas in phase 1 and possibly over 1 million in phase 2. When live, the radio telescope will generate 1 exabyte of raw data and over 1 petabyte of processed data every day. This is big data at scale and there is only one way to handle it efficiently and that is with HPC technologies.

HPC and Supercomputers have been around for decades and often perceived as being a niche IT offering reserved for large research labs or corporations due to their complexity to deploy and cost. It is fascinating to see that HPC has helped solve large problems such as understanding diseases better and finding cures for them, designing safer airplanes digitally or predicting extreme weather events more accurately. Beyond this, HPC has long been leveraged in more mundane examples to find the best curve of a crisp, to design Olympic record breaking swim suits or leak-proof diapers.

Today, HPC has really become pervasive across industries, a tool readily available to most organisations big and small and is no longer the turf of government labs, universities and large companies. But beyond the amazing processing capacity offered by HPC, I believe it offers a unique opportunity to manage efficiently the data deluge triggered by the ever more connected world we are living in. The long history and investment from the HPC industry to extract every last ounce of performance across all components of HPC architecture—be it memory, compute, storage or networking makes it uniquely positioned to facilitate and handle the required speeds and feeds of an impatient world.

No one is better positioned than an HPC expert to architect a complex solution that will seamlessly move data from the Edge to the Core, from memory to compute and then to storage, whether on premise or in the cloud and all this leveraging low latency networks and next generation wireless technologies. Most of all, nobody understands better how valuable data has become and how critical it is to be available in real time, anytime. We live in a data centric world where we need answers fast to handle the 3.5 billion Google searches per day, the 400 hours of video being added to YouTube every hour or the 31 million messages posted on Facebook every minute.

Data is the new IP and its value is derived from our ability to turn data into insight, insight into action and action into success. Finding valuable learnings from millions of sources and data types has become the new IT frontier and to predicate how data will be used to augment human capabilities and performance through machine learning and artificial intelligence.

All of this is why HPC matters.