Scale up to on-demand, High Performance Compute with HPCFLOW

Posted: 24 Jul 2018

From today’s big data to CAE to AI and machine learning workloads, high performance computing (HPC) is in high demand. But some companies’ and organizations’ budgets do not have the CapEx available to deploy a new HPC cluster — or to add HPC resources to their already burdened HPC cluster.

Advania Data Centers (ADC), a leading Nordic HPC, data and cloud service provider, presents HPCFLOW — an HPC as a Service (HPCaaS) offering that enables customers flexible access to HPC resources. They pay only for the resources they use. ADC’s team of highly experienced HPC engineers and architects work behind the scenes to provide HPCFLOW customers supercharged productivity 24/7 with no queue to execute HPC jobs.

“On-demand or consumable HPC is evolving at the moment,” says ADC’s Technical Director, Jón Þór Kristinsson. “HPCFLOW is a service where you can host your HPC environment and access resources in an elastic manner that gives you flexibility and secure HPC operations. We have the technical capabilities to serve any HPC operator with state-of-the-art infrastructure that is easily accessible and scalable. We also accommodate the business needs of HPC operators and make the business of the HPC operation more attractive than the traditional way of acquiring your own HPC cluster.”

HPCFLOW‘s technology stack (see figure) is based on OpenStack, a proven solution for cloud computing. Depending on the customer’s needs, provisioning is available for both bare metal HPC and virtual machines. We also have a range of HPC software solutions available – including ANSYS and Numeca – with no setup, tuning or tweaking needed (SaaS). For storage, HPCFLOW offers both central storage with tenant isolation and object storage for long-term data, all within its HPE-powered high performance computing clusters.

HPCFLOW‘s high-speed interconnect is based on Intel’s Omni-Path Architecture with tenant isolation and features 100Gbit connectivity for HPC workloads. Clients accustomed to InfiniBand, with software libraries based around Mellanox architecture, can however rest assured that their software will also run with ease on HPCFLOW.

Customers who already have HPC infrastructure of their own still may need rapid scale-up and scale-down of additional HPC resources. For these customers, HPCFLOW also provides an attractive solution.

“People can easily build on existing HPC infrastructure that they already have,” says Staffan Hansson, HPC Sales Specialist at ADC. “HPCFLOW will work with VPN connectivity to provide a seamless operating environment that includes your existing HPC resources.”

Manage HPC Resources with HPCFLOW’s Self-Service Portal

Especially when accessing cloud-based HPC resources, hands-on operational management of the system is essential. HPCFLOW satisfies this crucial demand with its self-service portal. The portal provides flexible user and project management — including network creation and management and bare metal and VM provisioning. Multiple users are allowed per project, enabling each user to have their own separate account. User quotas are also available for servers and storage. And security group management is enabled for network isolation.

“The self-service portal allows the users to provision their clusters to create a full solution for HPC,” says Jón Þór Kristinsson, Technical Director. “This includes provisioning of compute nodes for bare metal and VMs, object storage and network file system (NFS) shares. The customer can also use object storage to fully encrypt his data. There are a lot of solutions in the self-service portal that the customer can use to create the solution that meets their individual needs for HPC.”

For customers that have already have their own cluster management software, HPCFLOW also makes it possible for you to use your existing software to manage the resources you have in our cloud via a VPN connection, thus boosting convenience and the simplicity of integration.

Automate HPCFLOW Operation with APIs – and Know We are Here to Help

As an added feature for more advanced HPCFLOW users, ADC makes application programming interfaces (APIs) available for the customer who wants to automate their remote HPC experience and schedule workloads.

“If I need HPC resources on-demand, it can open up a totally new set of possibilities for me if I can automate my usage with APIs,” says Jón Þór. “If I know there’s a workload that I run every day, week, month or year, I can create a script for that workload that will spin up resources and deploy the workload in HPCFLOW. That could save me hours of work every time I need to run it.”

While the self-service portal and APIs allow customers to direct their resources independently, ADC recognizes that there are times when personalized support is needed. Our team of HPC experts are on-hand to help around the clock with issues surrounding cluster creation, set up and management. This includes performance validation and tuning assistance as well as applications support.

Put Security at the Forefront

ADC understands that our HPC customers are often dealing with sensitive data and that security is of the utmost importance. To address these concerns, we ensure that compute nodes are reserved exclusively for each customer. Once projects are complete, we wipe drives before putting them to use for others. Furthermore, we offer a range of encryption to ensure secure communication. This includes encrypted data transfer and encrypted communications between all servers and auxiliary and computation nodes. If desired, nodes and data at rest can be encrypted as well.

See What We Can Do for You

As a flexible and scalable HPC solution for rapid adoption of high performance compute and storage, HPCFLOW represents a future-proofed step toward the next generation of HPC. It multiplies productivity with access to performance tuning and support when needed. Companies and organizations that deploy HPCFLOW secure a competitive edge while maintaining a lean HPC team.

ADC offers prospective HPCFLOW customers free benchmarks to discover how rapidly their workload and application stack speeds up running in a leading-edge HPC environment. To find out how HPCFLOW can change your compute and HPC experience, contact an ADC sales representative today or visit our dedicated HPCFLOW page.