UK’s largest academic cloud to meet a broad range of workloads, including HPC by Sardina Systems
One of the leading universities in the world, amongst top 6 in the UK, has recently implemented the largest academic cloud system in the UK which is used for HPC workloads as well. The upgrade phase happened with zero-downtime.
The popularity of private cloud adoption even in the academic sector is rapidly rising. Today, OpenStack in academia is not just a subject we discussed at summits or conferences anymore.
The University of Edinburgh decided in early 2016 to implement an OpenStack-based private cloud, to be able to provide the University’s researchers with more flexible self-served computing services. The University would like to build the solution on an advanced, flexible and cost-effective platform based on open source technologies. The OpenStack-based private cloud, conceptualized in April 2016, had to be operational before the commencement of the academic year in September 2016.
Thinking from the opportunities perspective, the system was designed to meet the requirements of a broad range of workloads, allowing HPC workloads to be catered to with maximum efficiency and performance and flexibility. The workloads include GridPP and CERN workloads. The researchers using the system are able to leverage the programmability of the FishOS cloud environment to automate workflow, to achieve a high dynamic scientific discovery process.
The objective of the OpenStack-based cloud, from a user’s point of view, was to enable additional and enhanced computational and data services tailored to the needs of the University’s high-value and fast-growing Data Science research workloads, including Spark and Hadoop computational workloads.
This computational cloud for research, now named Eleanor, provides a scalable computational cloud platform accessible directly by researchers, complementing the computational capacity of the University’s Eddie 3 Service.
The solution design was aimed at an initial size of 200+ hypervisors, to be horizontally scalable.
The OpenStack-based cloud includes a Ceph storage system designed to deliver high performance and reliability while balancing optimal cost and delivery timelines, with a balanced configuration of SSD and NL SATA drives set up for 3x replication. With distributed virtual routers (DVR), OpenStack networking was designed to achieve high performance with optimal cost, balancing off bottlenecks.
All the management services operate in highly available (HA) mode, to the extent that entire racks of management servers can be turned off without impacting the private cloud service, thus ensuring service uptime.
More info about the entire solution design that allowed to the University to achieve service-assuring Zero-Downtime Upgrade on our article on Cloud28+.
About University of Edinburgh
The University of Edinburgh is one of the leading universities in the UK, ranked 19th in the world, 6th in Europe and 4th in the UK. It is largest university in Scotland with an international reputation for innovative research — 83% of the University’s research classed as world-leading and internationally excellent. The University’s Information Services Group, comprising some 700 staff, serves the University’s 36,000 students and 13,000 staff.