UbuntuNet-Connect 2015 Conference Papers and Presentations

Permanent URI for this collection

https://africarxiv.ubuntunet.net/handle/1/198

Browse

Browsing UbuntuNet-Connect 2015 Conference Papers and Presentations by Author "Becker, Bruce"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Executable Infrastructure for Regional Collaboration
    (2015-11) Becker, Bruce
    E-Infrastructure has, over the years, proven its worth in enabling scientific collaboration, even at regional and global scales. The adoption of common platforms such as HPC, data management, etc has made sharing of scientific data, applications and research outputs more appealing and is accelerating scientific output, especially in regions where these were previously unavailable - particularly Africa, and areas of the Arab-speaking world. The CHAIN-REDS and ei4Africa projects have supported the development of the Africa-Arabia Regional Operations Centre (AAROC). The ROC acts as a coordination point, first for grid infrastructures, but has been expanded to more general collaboration infrastructure services such as science gateways, federated identity providers, document and data repositories, etc. This expansion of service offering to ever-more demanding research communities places unreasonable strain on a fully-distributed model, where every site administrator is expected to understand and operate these new services. This is particularly true in the African and Arabian regions, where knowledge networks are sparse. This contribution describes a development and deployment philosophy which adopts a "DevOps" paradigm which aims to encode models of services using Ansible, with Github and Slack as collaboration platforms. Site and service configuration has been coded into Ansible playbooks, providing a reproducible model of the service, which can be customised as desired on a per-site basis. Most importantly, this model is executable, meaning that any number of sites and services can be effectively deployed remotely, by a core team. Continuous integration is done at every commit of code, by executing the playbooks on a cloud-based development site, which provides transparency to the remote site administrators. The adoption of this methodology helps to solve the problems of sustainably maintaining service configuration, improving communication between site operations and service developers, ensuring the proper state of services, and verifying the state of deployment. Some of the main benefits of this approach are speeding up the deployment of new services, reliably applying updates and recovering from disaster. In this contribution, we show that the "traditional" HPC and grid service deployment can be reproduced and improved, by adopting a more modern operations stack. However, we also highlight how this has helped to deploy advanced services- federated identity infrastructure, science gateways, application repositories, and Open Access repositories in particular throughout the region in short time, and is now playing a crucial role in the strengthening of technical and scientific collaboration networks in the region.