Octoshell: Supercomputer Center Management System
The development of Octoshell system was inspired by MSU Supercomputing Center administrators who faced a strong need in powerful tool to manage several supercomputers simultaneously.
The key features of the Octoshell system are:
- easy user accounts management;
- flexible quotas control;
- effective users support system;
- immediate users and projects statistics for administrators and managers;
- no modification to system software;
- handy incorporation of new machines;
- expandable functionality.
Octoshell was designed to be highly expandable by use of custom modules, i.e. it should be able to handle monitoring data or to control supercomputers hardware as well. The important point of Octoshell is easy way to integrate with supercomputers. If Octoshell is turned off, supercomputer continues to work normally, and administrators are able to control it as usual.
Octoshell is targeted to support collaborative work of users. By default, each project corresponds to POSIX group, and all users in the project have personal logins. If user participates in several projects, he has separate logins for each project. This simplifies quotas management and generation of reports. In general, this approach proved to be useful in MSU Supercomputing Center. But it is possible to alter this behavior and implement another approach to users and groups management.
Dmitry N., Vladimir V., Sergey Z. Resolving frontier problems of mastering large-scale supercomputer complexes // Proceedings of the ACM International Conference on Computing Frontiers (CF'16). — ACM New York, NY, USA, 2016. — P. 349–352. DOI: 10.1145/2903150.2903481
Nikitenko D. A., Voevodin V. V., Zhumatiy S. A. Octoshell: Large supercomputer complex administration system // 1st Russian Conference on Supercomputing Days 2015, RuSCDays 2015; Moscow; Russian Federation; 28 September 2015 through 29 September 2015. — Vol. 1482 of CEUR Workshop Proceedings. — 2015. — P. 69–83.