As we prepare to descend upon Frankfurt this week for ISC18, science will take its appropriate place at center stage. Throughout the week, the gathering will hear how through science we are challenging the world’s toughest problems, dissecting those problems down to their foundations and then building them back up by methodically moving every minute detail that has been learned into the powerful realm of the supercomputer. Science shows us that the best innovations are created by literally starting from scratch and building from there.
In business, it is a rare opportunity to be able to challenge a problem in such a methodical and primary manner. As entrepreneurs we relish the opportunity to literally wipe the slate clean, to think, to design and then most importantly, to build. The team at Verne Global has been relishing every step of our journey to deliver hpcDIRECT, a TrueHPC bare metal computing cloud delivered at industrial scale. We are delivering hpcDIRECT to serve the emerging entrepreneurs of high performance computing (HPC). Our journey started with an idea, a ‘blueprint’, and a desire to create the optimal environment for the leading HPC applications. We are excited to bring that message to ISC18 this year.
Our idea begins with a secure data center campus that offers a range of solutions that are optimised to the high intensity applications that will drive the future of innovation. These applications require huge amounts of power, and so the story of hpcDIRECT starts with energy. We knew that in order to provide a TrueHPC solution, we had to build our products from the ground up, or in fact beneath the ground and up, when you consider the 100% sustainable energy resources that power our infrastructure. We started our journey towards TrueHPC by choosing a location for a data center campus that could depend upon ample power resources for decades to come. In Iceland, we not only found access to extremely reliable, 100% renewable energy, but we also found a fully established green energy ecosystem that has been built to serve large scale, power-intensive industries. Because of this modern and industrial scale ecosystem, we were able to secure long-term power pricing protection.
The next phase of our journey was into the physical data center itself. Truth be told, very few data centers in the world are purpose built to deliver TrueHPC performance. HPC applications are not only power intensive, but also vary widely in terms of their demands on the underlying infrastructure. By working with our customers who were deploying their own HPC infrastructure at scale, we developed physical data center designs that are flexible to meet the needs of today's HPC infrastructure. This flexibility is not only related to the power density of the equipment but also related to the resiliency requirements of the applications themselves. For example, on our campus we house HPC equipment for financial institutions that require 100% availability during all contingencies, including normal system maintenance. And on the same campus, under the same innovative design, we house equipment for manufacturing companies that use dynamic scheduling to adapt for scheduled infrastructure downtime allowing for reduced resiliency and therefore reduced costs to the end user. Our data centers are designed to naturally flex between these varied resiliency requirements and to adapt to a wide range of power densities, all while taking advantage of the cost savings associated with our unique free cooling designs uniquely available at our Icelandic campus.
Our data center campus in Iceland has been very successful for a variety of HPC applications. For research institutions such as Earlham Institute that are working with advanced computing on the next scientific breakthroughs; for financial services customers that are leveraging HPC resources for better strategy and more advanced methods of compliance and risk analysis; for manufacturers that are designing and testing the next tier of innovative products; and for forward thinking companies like DeepL who are delving deep into the next generation of machine learning, our TrueHPC data center solutions have provided our customers with access to industrial scale facilities and industrial scale access to the only 100% renewably powered grid on the planet. And built upon that industrial scale infrastructure, Verne Global now offers an on-demand bare metal as a service computing platform called hpcDIRECT.
hpcDIRECT is developed with all of the tenets of TrueHPC in mind. Since hpcDIRECT is designed from the outset to provide our users with 100% bare metal access, our cloud offers full accessibility to their compute resources and complete flexibility to customise their cluster, their powerful servers, their dedicated networks, and their secure storage for optimal application performance. Thanks to hpcDIRECT’s industry standard orchestration tools, our customers are empowered by the flexibility to federate their applications across clouds, ensuring that the goal of the optimised application begins with the optimised location. TrueHPC promotes federation, not cloud lock-in.
Choosing an optimised location requires a computing cluster with the right performance for the right HPC application. hpcDIRECT is designed and implemented to deliver accelerated performance. We provide a wide range of standard SKU’s as well as bespoke configurations that are provisioned with important high performance features. Our CPUs are Intel® Xeon® Scalable Processors, formerly Skylake, Gold 6130 and Gold 6154 as standard options with RAM loadouts up to 768GB as standard. Because our customers have a wide range of applications and an even wider range of storage requirements, we provide server options with NVMe or SSD drives as standard. Fast shared storage solutions are also available with multiple options, ranging from dedicated to multi-tenant, depending on the usage and security requirements. And each of the hpcDIRECT server options are provided with both conventional Ethernet and Infiniband EDR interconnects for high data flow rates across the compute nodes and fast storage networks. Finally, we provide server options for both NVIDIA® Tesla® P100 and V100 GPGPU cards with additional features specifically tuned to machine learning applications. The true performance of hpcDIRECT lies in our foundational ‘blueprint’ that provides a fully customised HPC cluster deployment with any variety of standard server, network, and storage options in minutes.
hpcDIRECT - STANDARD SERVER OPTIONS
Our products are designed and built to deliver full transparency. TrueHPC applications are built cloud native and we believe that a cloud native solution should not obscure the application manager’s view of the underlying infrastructure. In fact with our monitoring solutions, many of our customers find a higher degree of awareness with hpcDIRECT than they have ever had with their own equipment. And this transparency and awareness embraces one of the TrueHPC goals of total security. As a bare metal solution, hpcDIRECT provides our customers the exact security state of the machines that they depend on for delivering their applications. Through the use of hpcDIRECT ‘blueprints’ to automatically deploy the HPC clusters, security operations such as operating system and library patching can be easily executed and the results can be thoroughly audited. By choosing hpcDIRECT, our customers gain more than just powerful HPC infrastructure, they benefit from industry leading support from a first class HPC devops organisation at Verne Global. Our philosophy is to provide devops services to our customers as if we were a member of their delivery operations team, and this approach has received outstanding responses from our customers who are not accustomed to the level of service we provide. At Verne Global, we use the simplicity of devops to reduce complexity of HPC operations and we can guarantee security thanks to the structured control provided through bare metal deployment.
TrueHPC is about true ownership. As we will see at the many presentations and panels at ISC18, high performance computing is mainstream and garnering both attention and scrutiny from the business leaders. HPC applications are shifting from early stage research and development to real time business productivity. A shift from the back office to the critical path of profitability focuses attention on two primary concerns: scalability and sustainable delivery. If HPC is driving more revenue at a lower incremental cost than other business solutions, smart organisations will look for solutions that will be able to scale with accelerating growth and these organisations will search for partners that can deliver these scaled solutions in a sustainable manner. A TrueHPC solution must be built to deliver at industrial scale. By anchoring hpcDIRECT on our expansive campus that has been built from the ground up to delivery high performance computing solutions, hpcDIRECT delivers a sustainable approach to industrial scale. TrueHPC is powerful, agile, and efficient. hpcDIRECT is TrueHPC.
Note: I'm looking forward to presenting at ISC18 on Tuesday 26th June, 4.20pm CET in the Exhibitor Forum. Please come along if you're interested in the concept of TrueHPC to hear more. Alternatively, come to our stand (F-912) and I'd be delighted to meet you.