Speaker
Description
Summary
Today's research in various fields such as physics, chemistry and biology have shows large demands in computational resources due to the complexity of tasks performed. Such resources are often provided as supercomputers and clusters for High Performance Computing (HPC).
The problem is that such platforms may have a high variance in their load. So more popular platforms may have large queues and significant waiting times, while other platforms are vacant.
On-demand clouds could help solving this problem by offering virtualized resources customized for specific purposes. Such clouds offer more flexible and convenient platform for researchers to execute their computational tasks, but their heterogeneity makes it hard to switch between platforms if some platform becomes highly loaded or inaccessible.
This happens because various resource providers have different interfaces (APIs) for task submission. So in order to change the target platform researchers need to spend time and resources to adjust their software for the new API.
In this research we propose MC2E - an environment for multidisciplinary academic research that aggregates heterogeneous resources such as private/public clouds, HPC clusters and supercomputers under a unified easy-to-use interface.
Comparing with “traditional” resource orchestration in data centers, that use free tools like OpenStack or commercial provided by VMware, MC2E offers a number of new features/opportunities and advantages:
-
MC2E provides higher level of resource control (a set of platforms instead of a single local data center or HPC cluster);
-
It provides users with more flexible capabilities to define virtual environments, more types of resources and services;
-
It supplies higher quality of resource scheduling and utilization;
-
It relieves a user from tedious system administration tasks;
-
It specifies a unified way to describe and support the data center (or HPC cluster) service livecycle. Allowing an MC2E user to use his software for providing experiments on MC2E infrastructure.
MC2E enlarges the concepts of PaaS and IaaS to the scientific applications area. We believe that it could be of great help to research groups that work jointly and need a shared virtual collaboration environment with resources from different geographically distributed platforms.
