HPE Pointnext Services experts can help identify cluster and resource management software that fits your high-performance compute use cases. Find out how to get the most value from leading solutions.
It’s the same in the high performance computing (HPC) domain. Cloud-native computing technologies are commonly used to provide greater agility, flexibility, and ease of operation by reshaping traditional architectures to Kubernetes-based and microservice-based architectures. Hewlett Packard Enterprise is a leader in the HPC market in the IT industry, contributing its technical development and delivery based on customers’ actual technical and operational requirements.
Through actual customer engagements in the HPC space leveraging cloud-native computing technologies such as Kubernetes, we see many companies facing challenges around the operation of cloud-native computing. The technologies are easy to use for simple workload configuration and deployment. However, they greatly increase the complexity of IT platforms for operations such as version-up, data protection, multi-tenancy, multi-cluster, high availability management, and disaster recovery.
Complexity drives cost of initial investment, soaring operational costs
This complexity often causes high initial costs and soaring operational costs, which results in many organizations giving up the use of cloud-native computing technologies. The Cloud Everywhere Practice team of HPE Advisory & Professional Services works together with our customers to address these challenges based on our long experience of HPC platform design/implementation and operation through our advisory, build, and operation services.
One of the key components, from the architecture point of view, is the cloud and cluster management function. Today there is a variety of software products available in the market that provide this functionality, such as SUSE Rancher, Morpheus, NVIDIA Bright Cluster Manager, and so on. All of them are great solutions, but which one is the best for your HPC use cases?
To make the best selection out of many choices for HPC platforms, it is critical to understand the origin and characteristics of the software and match them against your business and technical requirements. In this article, I describe three of the most popular cluster and resource managers used in HPC platforms and their typical capabilities. I’ll cover NVIDIA Bright Cluster Manager, Red Hat OpenShift, and SUSE Rancher Kubernetes.
NVIDIA Bright Cluster Manager is software developed by Bright Computing Inc. (acquired by NVIDIA last year) that can manage servers, Kubernetes, and other cluster infrastructure and resources for the enterprise. The Bright Cluster Manager for HPC provides efficient management and operation, strong UI and CLI, flexible deployment, and HPC-related software integration.
- Efficient management and operation. The software enables efficient management and operation of heterogeneous clusters in various environments, including hybrid cloud, multi-cloud, on-premises, data center, and even edge environments.
- Strong UI and CLI. NVIDIA Bright Cluster Manager provides an easy-to-use user interface, web dashboard and CLI command tools. Many cluster software solutions lack a friendly UI and CLI, which makes everyday operations difficult.
- The ability and flexibility to easily deploy and manage clusters of various sizes, from a few nodes to several hundred thousand nodes, using pre-defined automation technology. This advantage extends to other computing resources, such as Kubernetes, OpenStack, Ceph, and Jupyter Notebook.
- Strong integration and configuration capability with a variety of HPC-related software and workloads, leveraging resources such as graphics processing units (GPUs). This is perhaps the most notable advantage of Bright Cluster Manager. From the software integration capability point of view, it has integration capability with Slurm, Altair PBS Professional, IBM LSF, Univa Grid Engine, Open Grid Scheduler (OGS), TORQUE, OpenLava, and many other workload managers in the HPC space. Additionally, it’s a great advantage in the HPC space to be able to use Singularity as a container runtime. Bright Cluster Manager has a long history in the HPC space, which explains many of its advantages here.
Red Hat OpenShift
Red Hat OpenShift is the most popular Kubernetes software, used by many enterprises in the HPC space. Red Hat OpenShift is based on Kubernetes and supports build and operation of resources, clusters, and workloads, in addition to the integration of third-party software for cloud-native application development, and operation as a consolidated cloud-native computing platform. It has strong advantages around DevSecOps and automation capability by integrating with Red Hat Ansible Automation. One challenge is to manage and operate non-Kubernetes technologies, since it manages resources through kubectl. Of course, Bright Cluster Manager can be integrated with Red Hat OpenShift to leverage advantages of both.
SUSE Rancher Kubernetes
SUSE has been working on some very innovative and unique developments such Longhorn, which is cloud-native distributed storage, and Harvester, which is an HCI platform handling virtual machines and containers in Kubernetes environments. SUSE Rancher excels at handling multiple Kubernetes clusters, allowing management of multiple clusters from a single, simple dashboard. Also, SUSE Rancher not only manages the Kubernetes cluster, but also includes the management of the software running on top of it. It equips various software catalogs and provides construction and management for them. In addition, SUSE Rancher has functionality to import and manage an existing Kubernetes environment, and SUSE Rancher can also be used in combination with Bright Cluster Manager.
HPE services: Cloud-native expertise for HPC
There are many more cluster management software solutions in the market providing a variety of functionalities. The point here is that it doesn’t make much sense to compare the characteristics, functionalities, and background of these software products without understanding your use cases. It’s more important to build an optimal solution stack for your use cases.
In this article, the features, benefits, and strengths of three cluster and resource managers were covered. NVIDIA Bright Cluster Manager seems to fit many usage scenarios, so it’s one to consider for your HPC use cases. HPE, Cray, and SGI, major players in the HPC space, have been working with Bright Cluster Manager for many years.
HPE can provide even more value here by helping you leverage the expertise, experience, and knowledge we have cultivated over many years.
Every enterprise has its own requirements on how traffic is to be managed for cluster management. HPE Advisory & Professional Services has strong expertise and experience backed by hundreds of project deliveries that have helped companies improve the overall customer experience with modern tools and infrastructure.
Learn more about HPE Pointnext Services – expertise, methodology and tools to accelerate and de-risk your digital transformation.
Services Experts
Hewlett Packard Enterprise
twitter.com/HPE_Pointnext
linkedin.com/showcase/hpe-pointnext-services/
hpe.com/pointnext
ServicesExperts
HPE Services Team experts share their insights on the topics and technologies that matter most for your business.
