On Tuesday at SC11 in Seattle, Adaptive Computing launched a new version of its Moab HPC Suite targeted at the needs of enterprise HPC users. Moab HPC Suite – Enterprise Edition was a logical next step for the company’s HPC workload management product line. At the heart of the solution lies the Moab intelligence engine, which self-optimizes the productivity, workload uptime, and meeting of SLAs and priorities for HPC systems as well as the HPC cloud.
In an official statement, Rob Clyde, CEO of Adaptive Computing, explained the impetus for the project: “It’s costly and complex to manage heterogeneous systems. Today’s clusters have mixed processors, GPGPUs, operating systems, and middleware. Managing all this complexity along with power and business priorities is our forte.”
Adaptive’s Vice President of Marketing Chad Harrington, whom I met with at SC, painted a clear picture of the of Moab product progression and the positioning of this latest release. Like this publication, Adaptive lives at the intersection of HPC and cloud. They have a pure HPC product and also pure cloud offering (for less-technical, business workloads like Web servers), called Moab HPC Suite and Moab Cloud, respectively. Moab Enterprise can be thought of as connecting the dots between these two products, but with additional features targeted to the needs of enterprise users running HPC workloads.
Enterprise users are looking for systems that are simple to operate, reliable and as close to turnkey as possible. They may not be as cost-sensitive as academic users because getting the job done, and doing it on schedule, is the main priority. To that end, Adaptive claims this release brings together the key enterprise HPC capabilities into a single integrated product and services solution. The user interface allows operators to select whether they want to run their workload on a physical or virtual node, transparently to the user.
Moab Enterprise aims to provide enterprises with the following capabilities:
- Scalability to accelerate job response and throughput, including high throughput computing.
- Application-optimized allocation policies and provisioning to get more results out of existing heterogeneous resources and reduce capital costs.
- Heterogeneous cluster consolidation maximizes resource availability and administration efficiency by managing them as one cluster.
- Simplified HPC submission and control for users and administrators with job arrays, self-service submission portal, templates, and administrator dashboard.
- Self-optimizing scheduling packs workloads and backfills around priority jobs and reservations while balancing SLAs to efficiently use all available resources
- Advanced scheduling and management of GPGPUs for jobs to maximize their utilization including auto-detection, policy-based GPGPU scheduling and GPGPU metrics reporting.
- Workload-aware auto-power management reduces energy use and costs by 30-40 percent.
Like large-scale supercomputing clusters, cloud involves thousands of nodes and lots of people trying to use shared resources. However, one case is physical and the other is virtual. One of Adaptive’s customers, who wishes to remain unnamed, has tens of thousands of virtual machines in the cloud. Adaptive believes this is one of the largest deployments of its kind, perhaps the largest. As explained by Harrington, if you don’t have intelligent scheduling, you risk wasting some very expensive resources, but with an intelligent workload scheduler like Moab, a utilization rate of 95 percent can be achieved.
President and co-founder Michael Jackson emphasized the scalability of Moab, stating in a company announcement that “Moab is grid- and cloud-ready, compatible with existing infrastructure, and extensible to manage your environment as it grows and evolves to petaflop and beyond.”
In its pure HPC incarnation, Moab HPC Suite has already been deployed in many of the world’s top systems, about 40 percent of the top 100 systems, based on TOP500 list rankings.
As part of this this latest announcement, which comes on the heels of the company’s tenth anniversary, Adaptive also announced a rebranding of the Moab product line. This means that Moab Cluster Suite will now be called Moab HPC Suite – Basic Edition, with the moniker ‘Moab HPC Suite’ becoming the umbrella label for a portfolio of products, including Moab HPC Suite – Enterprise Edition and Moab HPC Suite – Basic Edition.
Adaptive also revealed that they are working on a cloud bursting technology with HP that combines Adaptive’s Moab HPC Suite with HP’s Cluster Management Utility (CMU), a tool for the management of Linux-based nodes in HPC clusters. Still in development, the combined offering will allow users to leverage the benefits of cloud in a traditional HPC environment. It does that by combining the real-time monitoring, management and provisioning capabilities of HP’s CMU with the intelligent decision-making capabilities of Moab Adaptive HPC suite.
Adaptive is currently giving demonstrations of this tool at their SC11 booth. The simulation shows a Moab Viewpoint interface with jobs running on a local cluster. When the jobs go above a certain number, in this case that variable was set to 50, any additional jobs trigger the provisioning of a virtual cluster in HP’s datacenter.