Verne Global

Data Center | HPC | Insights |

8 December 2017

hpcDIRECT - Powerful, Agile, Efficient

Written by Tate Cantrell

Tate is Verne Global's CTO and is responsible for setting the technical direction for the company. You can follow Tate on Twitter: @tate8tech

At Verne Global we are driven to create solutions. When we look to create solutions, we always look first to the needs of our customers and partners. Not only do we make it a priority, we feel it an obligation to communicate with our customers and learn where the industry can be challenged. At Verne Global, we love the prospect of meeting a new challenge and carving out a successful new solution.

This week, we announced the release of a new High Performance Compute as a Service built on bare metal compute nodes. We have named Verne Global’s new product hpcDIRECT, which is a direct response to the challenges that our customers are facing as they look to leverage their market position with more advanced forms of research and computational analytics. And, considering the trends in enterprise compute, our HPC as a Service product will be a solution for a growing number of organisations. The Next Platform said it best just this week: “Down the road, we think a much bigger transition is coming, one where Enterprise Computing Platform itself has a bare metal flavor for those who don’t want anything running in virtual machines, who think of containers, not hypervisors, as the smallest unit of compute.”

When we first spoke with our customers about the possibility of HPC as a Service, the main driver was to reduce the time from a user’s internal requirement to the time when that user’s HPC application was fully served and scaled out. When we asked where could we provide a solution, our customers requested the same powerful solution that we offer with our highly acclaimed colocation services, but they wanted this solution delivered at a higher level to streamline procurement, to save money, and to further simplify ongoing operations. They wanted maximum computational power from bare metal servers, while taking advantage of a financial shift from capital expenditure on the balance sheet to a more predictable operating expense line.

In response to our customers we set out to implement hpcDIRECT in the same way as we did when we brought the world’s first 100% renewably powered data center campus to the market in 2012. With our data center campus, we didn’t start at the top and work our way down, we started at the foundations of compute by establishing robust connections into the powerful Icelandic transmission systems and we ensured international service options by directly connecting to the sub-sea cable systems that strategically connect Iceland with Europe and North America. In just the same vein, we have focused first on the foundations that make up our innovative HPC bare metal as a service platform. We have designed hpcDIRECT from the ground up to ensure that with hpcDIRECT, our customers have visibility not only into their applications and into the data center and the underlying infrastructure, but also direct control into the orchestration of our bare metal service offering. We have designed hpcDIRECT to ensure that our customers’ dedicated servers are even more accessible than their own servers with technological innovations that challenge the status quo in the industry.

hpcDIRECT is powerful because it is designed and implemented from the outset to deliver bare metal servers to customers that want the full compute capacity of their high performance compute nodes. With hpcDIRECT, our customers have the ability to not only provision their bare metal compute clusters via a set of industry standard API tools, but they also have the ability to execute changes to BIOS settings just as they would with their own dedicated equipment. We have accomplished both power and agility through our bespoke implementation of OpenStack’s bare metal orchestration tools. With our bare metal as a service solution, our customers can automatically deploy computing clusters with infrastructure as code templates such as Heat Orchestration Template (HOT) and CloudFormation, or they can work through a dedicated and secure customer portal environment to design, deploy and monitor their applications.

At its core, hpcDIRECT is efficient by design and by location. Our bare metal servers are designed to ensure that no overhead is wasted on orchestration and every cycle of compute is delivered to our customers’ HPC workloads. We are efficient by location because all of the hpcDIRECT compute clusters are located inside powerADVANCE, our highly acclaimed data center product. powerADVANCE offers 100% uptime guarantees, 100% renewable power and 100% year round free cooling for all of our hpcDIRECT compute nodes.

But hpcDIRECT is not just about efficiency, it is about matching compute design with data center design to ensure maximum availability to HPC applications. powerADVANCE matches perfectly with our triple redundant hpcDIRECT orchestration platform. Since powerADVANCE offers up to four completely isolated electrical systems within a single data center deployment, we can naturally spread our orchestration platform across three independent electrical systems to ensure a higher level of service delivery than a standard cloud service implementation. This focus on service and dependability gives our customers confidence to extend their deployments beyond our world class data center services and into the very servers that make up their HPC footprint.

As you would expect, our customers have demanded the ability to deploy their HPC clusters with not only the highest level of compute but also with the most advanced networking options for their HPC workloads. With hpcDIRECT we offer the latest Intel Xeon (Skylake) multi-core processors with high throughput Ethernet networks for North-South connections and even faster Infiniband interconnects for East-West networking standards such as MPI. Because of the design of our orchestration control plane, we guarantee network isolation for all of our customers with dedicated port and Layer 2 segregation on Ethernet networks and dedicated port and subnet isolation for their Infiniband networks. This allows our customers the ability to draw up the customised compute cluster that they require to meet the specific needs of their HPC applications.hpcDIRECT is powerful, agile and efficient. And hpcDIRECT is designed, built and operated by a dedicated team of HPC professionals. Verne Global is proud of the success that we have helped to create for our diverse data center customers and we are excited to bring hpcDIRECT to market. With hpcDIRECT, our customers will find it even easier to deploy HPC workloads in the optimal location - a location with access to ample supplies of clean, renewable energy sources. As we like to say, innovation will come to those who have access to industrial scale compute. Verne Global and hpcDIRECT are built from the outset to achieve industrial scale compute.

Interested to find out more about hpcDIRECT? Click here. Next Platform's full article on hpcDIRECT is also available here.


Sign up for the Verne Global newsletter

Opinion, thought leadership and news delivered directly to your inbox once a month.