Why SoftwareOne is just the ticket for Entur

A project to improve scalability and automation using the Autopilot feature on Google Cloud Platform

The travel planner Entur was launched in 2016 as part of a government initiative to centralize all transport information in Norway and to provide a shared ticketing system for the train companies operating in an increasingly privatized railway network.

“The competing companies are private, but Entur is owned and run by the Norwegian government,” explains Tor Magnus Castberg, Group Lead of Shared Services at Entur. “To get around the problem of fragmentation, it was made a requirement that all rail companies use the same ticketing system.

They can have their own apps and web pages in the frontend, but they must use the Entur backend for journey planning, ticketing, and payment.”

The biggest rail operators are Vy, which runs local, intercity and regional trains on 15 different train lines in Norway including in the Oslo conurbation, in the east of the country and the busy Bergen regional line; SJ Nord which won contracts for seven lines including the Rauma and Røros routes; and Go Ahead Nordic which operates services on the Arendal, Jæren and Sørlandet Lines.

“What’s interesting is that SJ Nord decided last year not to develop its own app for its Norwegian rail business,” says Castberg. “They just send everyone to the Entur app instead.

“Businesses typically want to have their branding, and ownership of their customers, but many Norwegian transport companies are not running their own app because they have decided it’s a cost they don’t need.

“It shows that Entur is a complete and comprehensive transport planner and a technological pillar of the Norwegian transport system, as was the intention.”

In October 2022, Entur engaged SoftwareOne as its Google Cloud Platform (GCP) technology partner.

“The relationship with our previous Google partner had run its course in more ways than one,” says Castberg. “Our contract had come up for renewal, and we needed a partner that could offer us more than reseller deals.

“We had quite a few bids going into the process. We were already using SoftwareOne to optimize our extensive license portfolio, but more than that we felt SoftwareOne had the profile and expertise to get us to the next level technologically.”

At launch, Entur had “inherited” an on-site legacy system from Norwegian State Railways (NSB) which Castberg and his small team immediately set about shifting to GCP.

“Basically, it’s been a full renovation of the entire system,” says Castberg.

“Entur is a publicly owned company, and you might think that we’d be quite traditional in our approach. But that’s never been the case. From the outset, we embraced the agile way of working, with a ‘move fast, break things’ mentality, which makes us a fairly advanced Google customer.

“I don’t think we took shortcuts, but our premise had to be to make things work, and not necessarily do things right at every point.

“And so we looked into several ways where SoftwareOne could help us fix some of the issues around us having moved so fast.”

Entur and SoftwareOne landed on a project to improve scalability and automation using the Autopilot feature on GCP.

“We do have autoscaling in place already,” explains Castberg. “We have a low and high watermark, and we scale within these limits. The bigger problem is selecting the right kind of node.

“You can choose very powerful nodes, or you can use very slim nodes. A slim node will be cheaper, but not as efficient, while a large node would be more efficient, but also more expensive.

“What you want preferably is a selection criterion or some kind of logic to make sure that node selection is optimized between the two. You don’t want to spend too much money, but you also don't want an inefficient node.

“This is the SoftwareOne project in a nutshell: finding that sweet spot between cost and performance – and automating that.”

Three months after project kick-off – and a few weeks before the Christmas break – SoftwareOne was able to demo the new capability using the Open Trip Planner app available on Docker. The demo highlighted the following aspects of the project:

• The minimal operational activity required to create new environments on GCP
• How Autopilot dynamically adjusts capacity based on the application’s needs, leading to more efficient resource utilization and cost savings
• How the application’s Docker image will be used in the GCP environment and deployed to the Google Kubernetes Engine cluster, emphasizing the seamless integration between application deployment and the cloud infrastructure.

“We haven’t taken any of this into our production environment,” says Castberg. “To repeat, the goal was to look at the utilization of the nodes and how we can improve it. We have upscaling in place, but it is not efficient today. Looking at the new technology with Google Autopilot as developed by SoftwareOne it showed there is real possibility there, but it was a small MVP, and we will want to run tests at larger scales.

“We run a fairly slim crew as I mentioned so we shall probably be implementing this in Q2,” continues Castberg. “The process should be relatively straightforward – that was one of the principal objectives of the project. Assuming that the code set up, we would probably have to do a few bits of configuration and then fire off the automation scripts. That should be more or less it.