UTFSM, Valparaíso, Chile
University of Geneva, Geneva, Switzerland
DESY Zeuthen, Zeuthen, Germany
Humboldt University Berlin, Institut für Physik, Berlin, Germany
The ALMA Common Software (ACS) framework provides Bulk Data Transfer (BDT) service implementations that need to be updated for new projects that will use ACS, such as the Cherenkov Telescope Array (CTA) and other projects, with most cases having quite different requirements than ALMA. We propose a new open-source BDT service for ACS based on ZeroMQ, that meets CTA data transfer specifications while maintaining retro-compatibility with the closed-source solution used in ALMA. The service uses the push-pull pattern for data transfer, the publisher-subscriber pattern for data control, and Protocol Buffers for data serialization, having also the option to integrate other serialization options easily. Besides complying with ACS interface definition to be used by ACS components and clients, the service provide an independent API to be used outside the ACS framework. Our experiments show a good compromise between throughput and computational effort, suggesting that the service could scale up in terms of number of producers, number of consumers and network bandwidth.
UTFSM, Valparaíso, Chile
The ALMA Common Software (ACS) is a distributed framework used for control of astronomical observatories, which is built and deployed using roughly the same tools available at its design stage. Due to a shallow and rigid dependency management, the strong modularity principle of the framework cannot be exploited for packaging, installation and deployment. Moreover, life-cycle control of its components does not comply with standardized system-based mechanisms. These problems are shared by other instrument-based distributed systems. The new high-availability requirements of modern projects, such as the Cherenkov Telescope Array, tend to be implemented as new software features due to these problems, rather than using off-the-shelf and well-tested platform-based technologies. We present a general solution for high availability strongly-based on system services and proper packaging. We use RPM Packaging, oVirt and Docker as the infrastructure managers, Pacemaker as the software resource orchestrator and life-cycle process control through Systemd. A prototype for ACS was developed to handle its services and containers.