“The team is confident it can increase cloud cost savings by at least another 20%…”
The Home Office’s immigration technology department says it has cut its cloud costs by 40 percent, following several nasty “bill shock” events.
In a refreshingly frank appraisal, the Government Digital Service (GDS) said that over-reliance by the team’s 500 devs on costly on-demand cloud services was identified during an audit as the primary culprit for previously spiralling costs.
The department has since worked with its unnamed cloud vendor to use the excess capacity that exists in data centres, accessing “spot instances” at discounted rates instead. (This cost saving approach is not without risk as the cloud vendor can request that the resources be allocated to a premium user at any time.)
All of the major cloud providers off the ability to buy IaaS this way.
The Home Office notes that: “To mitigate the risk of losing vital computer capacity, Immigration Technology does not use the excess capacity option to power important production services that must always be available. But it is perfect for non-production services or temporary jobs, which currently make up about 30% of the compute powering their non-production containerised clusters.”
Turn the Lights Off on Your Way Out
The immigration technology team started to identify which services it did not use 24/7, such as processes and storage mechanisms that did not run at night time or over the weekend. Once a service that could be turned on and off was determine; the team began to schedule these processes, resulting in a 60 percent reduction in costs that had no impact on functionality.
Rather than suggest its teams use autoscaling features – which automatically adjust capacity use to suit processing needs – they just incorporated autoscaling as a standard component into all of their building templates.
The department notes that with regard to the Access UK online visa application service: “The team went from running 20 containers in production down to 2 containers. These scaled up only when there was an increase in demand.”
Home Office Cloud Costs Reduced
One of the key findings during the department’s investigation was that certain environments did not have the appropriate storage and compute resources allocated. When they looked at instances they saw that many of them were only using 10 to 20 percent of resources, assets that had already been paid for.
Pre-production and production environments had been set up in the same way to simplify development, the GDS found. However, this resulted in unused development services running as if they were a full production service.
To get around this issue the department set standard build templates to scale services appropriately to fit their purpose. In doing so: “Immigration Technology created a ‘small’ build template for development purposes, which does not include high availability options and keeps costs down. Immigration Technology uses a more expensive ‘large’ build for production services, which need high availability options.”
Many other cost saving measures were implemented; such as making some service cloud native, folding multiple databases into managed database as well as extensive containerisation which was implemented across the department to great effect.
The Home Office plans to inform teams in other departments about its cost saving measures and claims that: “By continuing these techniques, the team is confident it can increase cloud cost savings by at least another 20% as they continue to experiment.”