HPC-as-a-Service: innovation at your fingertips!
The high-performance computing (HPC) community has generally been slow to adopt cloud compared to other industry sectors for several well-documented reasons:
- Compromised performance due to virtualized compute instances
- Applications availability and difficulty with licensing
- The complexities of setting up an HPC environment within a public cloud framework
- Requirements for low latency interconnect and fast, parallel file systems
- Concerns over security
However, today many of these barriers to adoption have been removed or mitigated and according to industry analysts such as Hyperion and Intersect360 Research HPC cloud is now the fastest-growing segment of the HPC market.
Recent surveys of major HPC facilities report that over 90% of all HPC centers expect to use cloud services for at least part of their HPC service provision by 2026[1]. It is crucial to note though, that this is generally supplemental to on-premises resources rather than replacing them completely.
Some market verticals, such as life sciences, were early adopters of cloud and are leading the charge, although it is now true that all sectors are growing, moving increasing portions of their workload the cloud to:
- Reduce the size and associated costs of on-premises facilities
- Move non-time-critical jobs to the cloud releasing on-premises resources for more time-critical works
- Evaluate new technologies without the high cost of capital purchases
- Burst workloads to manage peaks and troughs
The Atos OneCloud initiative
Atos OneCloud is an industry-wide initiative to help organizations migrate workloads to the cloud, finding the right balance points that work for them. It is a ground-breaking initiative that blends cloud advisory consulting, application transformation expertise, prebuilt cloud accelerators, and innovative talents in an end-to-end set of services to help our clients navigate their cloud journey securely. Atos OneCloud modernizes enterprise applications to be agile, mobile, and rooted in analytics, across public, private, and hybrid cloud environments.
What does Atos OneCloud mean for HPC?
There are three key service components required to deliver a comprehensive user experience for HPC in the cloud:
- Simple point and click interface and API batch service endpoint for HPC provisioning, including bursting containerized workloads that can be expanded on demand on the target public cloud platform
- Job submission to cloud via API or traditional schedulers such as SLURM or LSF gateways with the ability to track usage and integrate this with an automated chargeback mechanism
- Fully converged digital transformation of HPC with Kubernetes to support unified, multi-cloud deployments
Enabling HPC-as-a-Service with the Atos Nimbix solutions
Atos’s recent acquisition of Nimbix[2] supplements its existing HPC cloud software environment and in doing so has created the industry’s most comprehensive hybrid HPC cloud portfolio. Nimbix JARVICETM XE was the world’s first container-native hybrid/multi-cloud HPC platform, enabling scale-out and accelerated HPC and AI workflows (SaaS and PaaS) globally on any Kubernetes compatible infrastructure.
Nimbix JARVICE XE brings industry-leading supercomputing cloud technology into the data center to support advanced computing workflows while maintaining the option to easily burst to any public cloud for additional resources. JARVICE XE uniquely delivers accelerated applications and workflows that take advantage of diverse infrastructure including InfiniBand, GPUs, and FPGAs on dedicated Kubernetes infrastructure. Specifically, it provides the:
- Ability to convert SLURM jobs into low level “Pod” scheduler jobs
- Flexibility to deploy scalable HPC jobs through containers
- Tenant isolation so that no clients share the same resources
- Nimbix HyperHub applications catalog to help the user define workflows
Nimbix HyperHub provides a multi-cloud open marketplace for HPC and AI applications.
JARVICE XE enables customers to easily burst from on-premises solutions to public cloud systems from Google, AWS, and Azure, and others or to manage internal systems as private clouds.
Cloud HPC Services
Of course, enabling access to resources is only one part of the equation. To provide a complete solution requires access to consultancy and professional services offerings to maximize the cost-effectiveness of the resources.
Through its global HPC Cloud Center of Competency and Center of Excellence in Advanced Computing, Atos is also able to offer clients additional services as cluster configuration services, workflow construction, HPC-as-a-Service portal customization, HPC application porting and optimization, HPC container development, HPC application benchmarking, HPC application analysis, and HPC Training services.
Atos can also offer additional bespoke services for specific customer domains, e.g. services for CAE simulation integration and Accelerated Genome sequence analysis.
Conclusion
HPC cloud usage is expanding across all industry sectors with a strong desire from customers for a flexible hybrid model that can be deployed in a multi-cloud scenario.
With Atos’ acquisition of Nimbix and the combination of Nimbix JARVICE XE with its existing tools and frameworks, Atos has created the industry’s most comprehensive portfolio of HPC cloud products and services with which to address the full range of envisaged HPC usage scenarios.
[1] Big Compute: 2021 State of Cloud HPC Report, HPC Practitioner Survey
[2] Atos Press Release July 27th, 2021: Atos strengthens its leading position in HPC cloud enablement and acquires leading global HPC cloud platform provider Nimbix
The post HPC-as-a-Service: innovation at your fingertips! appeared first on HPCwire.