🖊️ This blog post has been co-authored by Yash Lamba, Senior Cloud Engineer, Oracle.
Researchers who want to use commercial cloud computing in conjunction with on-premises resources for their research workloads often find it challenging for a variety of reasons, including the requirement to use different and unfamiliar interfaces to set up, schedule and submit workloads. With OpenNebula 5.12, it just got easier to use Oracle Cloud Infrastructure (OCI), because OpenNebula now supports cloudbursting to OCI.
OCI is Oracle’s next-generation cloud infrastructure that delivers high-performance computing power and services through a global network of Oracle public cloud regions to run cloud native and enterprise and research IT workloads. OCI provides high-performance computing, GPUs and real-time elasticity by combining Oracle’s autonomous services, integrated security, and serverless compute.
Did you know that with the new OpenNebula OCI driver, users can burst from on-premises resources to Oracle Cloud and seamlessly spin up VMs in OCI, extending the hybrid cloud experience to research communities? With the OpenNebula OCI Driver, you can spin up an instance using any available shape in OCI in any Oracle Cloud region of the world. The instance can use a GPU shape, a Dense IO shape or even a flexible shape based on the workload. The instance can belong to any existing network or can be a part of a different network depending on your needs. In fact, the machines can even run any operating system image supported on OCI, including custom images that you can export from your on-premises environment to OCI.
In short, cloudbursting extends the limits of on-premises computing clusters by enabling the dynamic deployment of applications and workloads that burst into public clouds when needed to meet capacity demands. This can be particularly important in research, and in any other context where computing resource consumption consistently varies significantly over the course of a project. It can be inefficient and prohibitively expensive for universities, companies, and other organizations to build on-premises clusters that support peak demand but otherwise are underutilized. A configuration that takes advantage of the vast capacity of commercial clouds when needed creates economic and infrastructure efficiencies. Are you into green computing? Don’t waste money, energy, and precious natural resources on supercomputers that will sit idle – cloudbursting is a better alternative.
While bursting from an on-prem environment into a public cloud for extra capacity sounds like it should be straightforward, in reality, it is a complex technical challenge. Research users, some of whom have little technical knowledge, need to be able to interface with the system to submit their workloads. Research workloads must be organized and prioritized within the on-prem environment. The on-prem environment must determine when and what type of extra capacity is needed and must be able to manage the workloads to optimize use of the cloud capacity. The research workloads that will move to the cloud must be transmitted and processed securely, efficiently and without disruption of the on-prem environment. Research results must be returned to the researcher in the same way – securely, efficiently and without disruption of the on-prem environment. Also, the cloud machines must automatically be spun up and spun down so research dollars are optimized.
Experts from several Oracle teams – Oracle for Research, the Oracle ISV and Cloud Native team, and the Oracle HPC organization – joined forces with Rice University to develop and contribute the OpenNebula driver specifically to support research workloads. Rice wanted to expand its on-campus HPC infrastructure with a cloudbursting solution that would also support the Open Science Group (OSG) and researchers across the United States. Oracle is committed to collaborating with researchers and accelerating research results with cloud computing. At Oracle, our technical teams engage in complex collaborative projects with customers every day, and Oracle for Research brings these technical collaborations to research teams at no cost. Working together, Rice and Oracle identified a system architecture and required capacity, and integrated tools like Kubernetes and SlateCI for containerizing and orchestrating research workloads, and connecting and aligning the on-prem architecture with OCI. The resulting driver was contributed by Oracle and Rice to the OpenNebula community.
Now we need you, the OpenNebula community, and your help. The OCI driver is available from the Community Add-ons repository, and we’d very much appreciate your help in testing it. Please check out the README file and the system requirements, then follow the instructions to install, configure and use the driver. Easy!
ℹ️ For more information about Oracle for Research and how you can gain access to Oracle Cloud Infrastructure and technical collaborations to advance your work, please visit https://www.oracle.com/oracle-for-research.