Annex 1 - case studies
National Libraries of Scotland
In 2007 NLS purchased two 100 TB Hitachi AMS1000 Storage Area Networks as part a project plan to ingest digital objects.
A pilot was set up to investigate the world of virtualisation by re-purposing two Dell 2950 servers to form a cluster running VMware Enterprise 3.5 utilising the new SAN as shared storage. Following a successful pilot a small number of general purpose servers were run from this platform.
VMware was not fully adopted in the Library until the original two Dell servers were replaced and relocated to our DR site running the VMware Site Recovery Manager solution with four powerful HP DL380 servers taking their place. And with upgrades to our tape library system the transition was made. All new server applications were deployed from VMware and a programme was embarked up to transfer the physical server estate into VMware accomplished without issue with minimal server downtime.
Today NLS have VMware ESXi host servers running over 150 virtual servers split into three clusters, Production, Disaster Recovery/Test Network and the NLS Digital Library System Node for the British Library each on their own SAN system. All manageable from a single location with the capability of failing over the most business critical elements to another building within an hour should disaster strike. All with the day to day benefits of manageability, responsiveness and reliability expected of IT systems today.
To sum up, the NLS production environment, excluding their DR, test/development and British Library environment, run ninety virtual machines [not including appliances that run VMware] on seven VMware ESXi physical hosts. With the capacity to run twenty virtual servers per host they can easily accommodate both their own and shared services partner's future growth. With the cost of a general purpose server for one application at around £3k vs. a server designed for our VMware implementation costing £13,500 this results in significant cost savings. Even factoring the cost of other equipment e.g. a SAN, the latest purchase was £42000 the virtual solution is vastly cheaper.
University of St. Andrews
The University of St Andrews underwent a data centre initiative to replace ageing legacy facilities with an efficient, reliable and scalable facility that would support its dedication to academic excellence for more than 8,000 students in around 160 buildings and with over 2,100 staff. At the time the University had 50 campus locations for servers with no fit for purpose server rooms.
A working party of internal stakeholders and external consulting engineers produced a business case for the construction of primary and secondary data centres in order to underpin reliable service delivery. At every stage of the process, energy efficiency was pushed and design decisions taken to minimize overall power consumption. Key parts of the design philosophy are free cooling, fully enclosed hot and cold air paths & ability to reuse the waste heat generated by the computing facility. The facility has a design for a full-load annualised PUE of 1.2. From an infrastructure perspective, the high efficiency of operation is underpinned by extensive telemetry of the datacentre, allowing optimisation of the equipment.
Running parallel with the population of the data centre, a path of increased virtualisation and use of shared and managed services to reduce the footprint was undertaken, with great success. The project delivered all the expected benefits and will deliver calculated savings against sector average of £1.4m over 10 years, a reduction of 6.8m kg CO2 emissions and is a stepping stone to the University's aspiration to Carbon Neutrality for Power.
Further improvement works are earmarked to optimise the PUE for part-load operations and reduce annual operating costs by a further £10K. The data centre has been awarded the BCS CEEDA Gold award.