Follow Datanami:
November 13, 2017

Workload Portability

Today, organizations of all sizes are exploring strategies for migrating workloads to and across diverse computing platforms.  The reasons vary from resolving local resource shortfalls to enabling collaborative workflows to converting capital expenses to operating expenses.  Whatever the motivation, the goal is simple: to make it possible to run workloads anywhere.  Recent developments in container and remote visualization technologies have greatly eased the challenges of delivering workload portability.

Containers are a key component of many migration strategies.  A container is a lightweight package encapsulating an application, its dependencies, configurations, and other resources required to execute the application consistently on a variety of infrastructures.  There are several popular container frameworks that continue to develop their capabilities.

Singularity is one such container framework that incorporates workload portability into its design.  Singularity gives the user full control of the environment, and can be used to encapsulate entire scientific workflows, including software, dependencies, and even data.  Singularity containers can leverage a host’s resources, such as GPUs and other accelerators, high-performance interconnects, file systems, and resource managers.  Moreover, this can be done without exposing elevated privileges.  This makes Singularity the obvious choice for scientific workloads, which may need to scale from a small development system to many thousands of hosts, such as in traditional high-performance computing environments.

Penguin Computing Scyld ClusterWare® is the popular HPC cluster management solution.  This complete HPC management solution enables super-fast cluster provisioning, guarantees configuration consistency through a single system image architecture, and provides easy-to-use management, monitoring, and job submission tools.  Scyld ClusterWare also fully supports Singularity containers.  This positions Scyld ClusterWare as an ideal platform for an on-premise HPC cluster environment.  Users can develop their workflows for cloud migration from the start.

Penguin Computing On-Demand™ (POD™) is a true HPC service in the cloud, designed and operated by HPC experts.  POD enables individuals and organizations to utilize high-performance, bare-metal HPC computing resources without having to invest in on-premise infrastructure.  And, POD supports Singularity, making it a superior option for hybrid cloud strategies.  POD also offers the other key technology that has eased cloud migration: high-performance, remote visualization.

Container technologies have enabled workloads to be scaled and migrated to the cloud with much more ease than ever before.  But, once a workload completes, what then?  Traditional HPC systems require users to download large results files to on-premise workstations for post-processing, a time-consuming process that makes it hard to create a predictable workflow.

Penguin Computing® Scyld Cloud Workstation™ is a remote desktop solution designed to address this challenge  Scyld Cloud Workstation delivers an engineering-class, real-time, interactive desktop, capable of utilizing graphical acceleration hardware, through a standard web browser.  The user is presented a standard desktop environment and can use the same graphical tools they would use on a local workstation, eliminating learning curve and boosting productivity.

When tightly coupled into a HPC environment, Scyld Cloud Workstation offers significant time savings by moving pre- and post-processing to a physical or virtual workstation with direct access to a cluster’s data storage.  This eliminates the need to download large results files and simplifies workflows.  Scyld Cloud Workstation is available for on-premise and cloud deployment, and is also featured on POD.

By pioneering emerging and open technologies, Penguin Computing enables our customers by delivering solutions that address real-world challenges, like workload portability.

Register for Penguin Computing’s upcoming whitepaper on enabling workload portability.