Genomics England Saves £1 million with FinOps from The Server Labs Using AWS

Genomics England saved over £1 million a year by adopting and implementing AWS Partner The Server Labs’ FinOps methodology to optimize and reduce the operational costs of its extensive use of Amazon Web Services (AWS). It uses AWS to support its genome sequencing work for diagnostics and research. Genomics England provides full genome sequencing for diagnostics for the UK National Health Service (NHS) and supports academic and commercial research. Systems must cope with 100 petabytes of data storage and constant demand for rapid diagnostics alongside variable demand from research projects.

Genomics England Limited (GEL) is a private company owned by the UK’s Department of Health and Social Services. It works with the NHS to provide whole genome sequencing for rare disease and cancer diagnostics. It also provides data and services to both academic and commercial researchers through its online platform the Research Environment based on AWS.

GEL is a major user of high performance and cloud computing, processing vast amounts of genomics and other sensitive clinical data. It sequences 80,000 whole genomes for the NHS a year which, along with research work, requires 100 petabytes of storage. Monthly NHS demand for GEL analysis has grown by 500 percent compared to the previous year. But systems must also deal with extremely varied demand from research projects which require high levels of compute capability for just a few days or weeks.

The need to manage larger and more varied datasets, along with the requirement to provide secure access to research partners, led to the decision to move to the cloud. In 2021, after considering various providers, it chose to migrate some key services to AWS for its security, reliability, and ‘burstability’—its ability to cope with fluctuating usage demands. To help meet strict and demanding timescales for the migration, GEL turned to AWS Partner The Server Labs (TSL).

While the migration has gone well, financial leaders were concerned about the visibility of spending and the risk of ‘bill shock’. GEL asked TSL to help it improve its visibility into its use of AWS services and help reduce costs by implementing its FinOps methodology. “We like and trust TSL,” says Katherine Stubbings, head of the office of chief information officer at GEL. “They understand the uniqueness of our set-up—we’re not a 24/7 ecommerce business, but we’re not a 9–5 corporate either. They’re the right size for us and can flex with us when needed, and have all the skills of a big player but are much more agile and available.”

GEL spending is spread across departments with fast-moving and varied technology requirements and systems. FinOps required a cultural change for engineers focused on solutions to improve services for diagnostics and researchers. The migration of hundreds of servers and hundreds of terabytes of data meant not all applications had been re-architected as planned.

TSL helped GEL visualize, understand, and manage its AWS costs and usage over time by using various tools such as a business intelligence (BI) dashboard based on Tableau Server on AWS linked to AWS Cost Explorer. This helped GEL find under-utilized or over-provisioned capabilities. Then, using TSL’s hierarchical FinOps methodology, it implemented serverless processes to automate SLA-based system runtimes. This included scheduled auto stop/start of test environments that were running 24/7. Switching off non-production Amazon Elastic Compute Cloud (Amazon EC2) and Amazon Relational Database Service (Amazon RDS) instances outside of working hours contributed 67 percent of total non-production savings of £1 million per year.

A key part of the FinOps methodology implemented by TSL was to provide FinOps ‘Golden Paths’ for the GEL engineers. These provided full documentation, together with example scenarios and code, so that moving forward, all platform engineering incorporated FinOps from the inception design. The BI dashboard and Golden Paths, together with increasing acceptance of FinOps has allowed GEL to hand responsibility back to individual departments, which in turn have been able to fund new projects and innovations by finding savings in their existing AWS usage. Now GEL teams are better able to match instances to workloads using Amazon EC2 for secure, resizable compute capacity in the cloud, and use Amazon RDS to remove inefficient and time-consuming database administrative tasks. GEL and TSL are still working on re-architecting applications and services to support a longer-term shift to microservices and alternative application architectures including more use of serverless technologies like AWS Lambda and AWS Fargate.

The teams are also seeking to increase efficiency and sustainability by more use of workloads running on AWS Graviton Processors. “The huge growth in our data means we have to be highly efficient just to keep functioning,” says Stubbings. “If we were still operating as we were a year ago, we could not afford to keep going, it’s a big challenge.”

Genomics England is very aware of its legal and ethical responsibilities to keep people’s most personal and sensitive information secure. The collection of processes that make up the pipeline from taking samples, storing, processing, and providing diagnostic advice are considered a medical device requiring ISO 13485 compliance and associated audits. “Genes don’t change—it is our most personal data and there’s an emotional consideration, too,” says Stubbings, “We take that responsibility seriously and appreciate sharing that responsibility with AWS.”

GEL originally reached out to TSL because of its experience with high performance computing and Big Data. In the coming years, GEL’s data volumes will increase dramatically due to its Newborns project and the Generation Study.

All babies born in the UK have a heel prick test—a blood spot test that checks for nine serious conditions that can be hard to spot and can benefit from early treatment if identified. GEL is piloting a project, called the Generation Study, to provide genetic testing of 100,000 babies to scan for 260 genetic conditions that would allow for intervention before symptoms of illness start to manifest.

If successful, the test could be used for 600,000 babies born in England and Wales each year compared to the 80,000 sequences carried out and stored now. GEL systems will need to keep growing to cope with this further increase in data and likely associated increases in research projects, something it can easily do with scalable AWS compute and storage services.

The Generation Study will also track the health of participants over their lifetimes, storing genome data for long periods of time and putting further demands on GEL’s systems.

“This is not a project that ends because we grow so fast,” says Stubbings. “We’ve saved £1 million but there are still savings to be made, where possible we’re looking at refactoring to use more AWS serverless systems. In the longer-term we may look to move storage from on premises to AWS.”