Case Study - Rehoming twenty years of Oracle, and the teams who ran it

Overview

The University of Sheffield is a Russell Group institution with more than 36,000 students, and at the centre of it sat a student record system that had quietly run everything, from first application to graduation to alumni, for over twenty years. It ran on Oracle: two ageing databases, Oracle Forms and Reports, Oracle APEX, and more than 560 applications and integrations threaded through 1,376 interconnected systems. Documentation was thin, the platforms were looked after by separate teams, and batch jobs ran long enough that starting one was a reason to go and put the kettle on.

Rather than refresh two decades of hardware and hope, the university chose AWS. Perform Partners led the programme, and brought Ologist in from August 2022. We stayed for three years, working across the whole of it: the analysis of an estate nobody fully remembered, the AWS build, the cutover, and the part we care most about - what the university's own teams could do once we left.

Twenty years of Oracle

Forms and Reports has been called legacy since before most of the university's students were born. It still ran the place, and technology that has given twenty years of service deserves a careful landing rather than a demolition. We uplifted both Oracle databases and rehomed Forms, Reports and APEX into AWS on patterns the old estate never had: auto-scaling, resilience by design, infrastructure defined as code. The processes stayed the same. The platform underneath them did not.

One long weekend

The cutover began on a Thursday night before a May bank holiday, well inside our first year. Testing started on Sunday, and on Monday morning everybody logged back in as if nothing had happened. The results were immediate: the reporting job that took 90 minutes now takes 6, fifteen times faster, and the university gained a disaster recovery story its old estate could never tell.

The teams who ran it

The platforms we migrated had been managed by different teams in different silos, and the estate showed it. So alongside the engineering we introduced the practices that make an estate one thing rather than several: source control, code review, infrastructure as code, devops as a working method rather than a job title. Then we made it stick. University engineers joined us in pairs, usually one from operations and one from development, working real tickets with an Ologist engineer alongside them. No classroom, no slideware; production work, guided. Between 50 and 60 of the university's 310-strong IT department came through it.

Three years on

We rolled off in August 2025. The work went on to win a Computing Cloud Excellence Award, and the university has kept moving without us: retiring its oldest applications, refining others for the cloud, exploring what a modern estate makes possible. You can read Perform Partners' account of the programme in their case study, and AWS published one of its own.

The migration took a year. Making it stick took two more. That is what we mean by Digital Permaculture: the job is not moving systems, it is leaving behind teams who can move their own.

What we did