This document describes the FLARECAST infrastructure architecture on a very high level. Featuring an agile development approach, we put our forces into writing readable and well structured code, rather than extensive documentation, which tends to become outdated too soon anyway. Consequently the main purpose of this document is to give a the reader hints where to look for detailed information.
Design Principles
- We designed and implemented a first prototype of the computing and development infrastructure. The main objective was to provide an infrastructure that can be installed on the FLARECAST computing hardware in UPSud as well as on the developer’s local machines. Obviously the local machines can host a subset of the full installation only.
Design approach
After analysis of the HESPE infrastructure we identified the concepts and parts to be reused for FLARECAST. Partially already reused is HESPE’s sophisticated concept for parallel processing of IDL routines in a cluster. In addition, some of the HESPE visualization tools will be re-evaluated at a later stage of the project.
For the overall implementation of FLARECAST we decided to take a different route. The main reason is that HESPE was designed for a closed group of developers and system administrators, while FLARECAST has pledged a plugin system even for external colleagues (ref proposal). Moreover, between the kick-off of HESPE, few years ago, and today, there has been a shift from cluster computing to cloud computing (ref).
The document contains the following parts:
- Revised version of the initial architecture model as outlined in the proposal.
- Implementation of the architecture based on a system of Docker containers.
- Deployment of the system in Paris.
- Documentation of the organisation of the FLARECAST source code repository