MIKELANGELO offers flexible, easy to maintain, open source, standardised and user friendly Cloud/HPC Cloud technology stack, focusing on performance, low footprint and security. MIKELANGELO components are easy to deploy and compatible with renowned upstream or enabling open source projects and standards such as KVM, POSIX, OpenStack and Kubernetes.
MIKELANGELO is addressing basically any HPC or Cloud provider (public or private), who have identified the need for an IO efficient virtualisation stack, combined with approaches optimised for performance of guest and host execution and communication. HPC virtualisation is impossible without high IO efficiency and very thin guest OS, thus MIKELANGELO is a direct enabler. In Cloud area, only specific workloads benefit from such efficiency, mainly because IO is of secondary importance in the Cloud. Moving closer to serverless architectures, this will become significantly important for Clouds as well.
MIKELANGELO solves the problem of virtualised IO in HPC and Cloud areas. More importantly, MIKELANGELO is offering a widely compatible and user-friendly set of approaches and tools, that helps with the IO optimisation, independently of the target infrastructure. With MIKELANGELO, we are employing a mix and match approach, where you can take specific components and integrate them into your specific infrastructure. For example, we offer unikernel technology (POSIX compliant OSv), with additional set-up tools; We offer IO optimisation technologies on the KVM (host) level - ZeCoRX and IOcm components and also between guest and host, with virtual RDMA. These components can be used on their own or all together. The same approach is with their integration - we can use them as-is, or integrate them into e.g. OpenStack or use them in GCE or AWS. Finally, we offer strengthened host security (SCAM component). Using MIKELANGELO components enables new virtualised architecture approaches and increased performance and scalability of the infrastructure as a whole. MIKELANGELO is also offering a range of supporting approaches and technologies, making the integration and use very straightforward and efficient.
MIKELANGELO consists of a set of technologies and approaches, all aiming to improve IO performance or to help with integration of MIKELANGELO components into larger systems.
With MIKELANGELO we can use unikernel instead of traditional guest operating system in virtualisation. This improves the efficiency, as unikernel is typically very small (50 megabytes), can be loaded in under a second and executes the workload as efficiently as standard guest (eg. Linux). On the host level, using approaches as IOcm and ZeCoRX (KVM modules), we can control the IO performance of the overall virtualised infrastructure on that specific host. We can even improve the guest-host communication or guest-guest communication under the same host, using the virtualised RDMA approach (if hardware support is available). It is important to note that OSv unikernel offers POSIX compatibility, however when such compatibility is not required, we can use novel programming approach called Seastar. Using such approach, we can obtain very high performance improvements, however we need to rewrite the application from scratch. Other approaches in MIKELANGELO are Scotty, for scientific experiments deployment and control, snap for highly efficient logging and telemetry, SCAM for strengthened host security, vTorque for deployment of virtual machines under Torque system, etc.
We expect business impact of MIKELANGELO to be on HPC Centers, integrators and owners of private clouds. These are going to benefit from flexible, easy to maintain, open source, standardised Cloud/HPC Cloud technology stack, which delivers performance, low footprint and has increased security.
Turning to the delivery of applications in the HPC/HPC cloud area, the packaging of typical HPC applications (e.g. OpenFOAM) into OSv packages, and ensuring their standardized management, should have significant impact for infrastructure providers that will be relieved from having to maintain multiple application versions specific to their clusters, as well as for infrastructure users that will be able to build and deploy applications on top of any MIKELANGELO-compatible infrastructure.
Furthermore, expanding the typical HPC applications to cover typical Big Data applications, will have impact in the High Performance Data Analytics field. Overall, we expect the contributions of MIKELANGELO to be accepted upstream into KVM community, thus delivering IO performance to the whole KVM community. The same holds for OSv, snap communities, where we expect all contributions to be accepted upstream, thus delivering value to all of the relevant communities.
Open Source software available on