TomTom takes the optimal route to the cloud with Microsoft Azure Storage solutions

Making maps with data

TomTom pioneered the cars that drive through our neighborhoods collecting street-view images. Years of this practice have yielded a repository of more than 17 petabytes (PB) of mobile mapping data, which continues to grow every day. On average, TomTom collects 3 million kilometers annually, representing 375 million panoramas or up to 2 PB of data. “That’s the starting point,” affirms Koen Denecker, Vice President of Platform Engineering at TomTom. “Imagine a digital image recorder of 3 billion images that support lidar data. Extracting what we call road furniture and the topology—the map itself—is a gigantic compute task that must be performed every time we add new features.” (Lidar is a 3D laser scanning technology used to measure distances.) And the company must distill the dense body of data from raw PBs down to a gigabytes-sized format for access by everyday devices in the hands—and cars—of consumers.

The company uses its photos and satellite imagery of the earth to depict a detailed topography. It processes mobile location analytics (MLA) data, translating the results into routing and hazard awareness information that it sends to mapping app users. Location technology companies like TomTom face the challenge of storing vast amounts of MLA and other location data—and providing the enormous computing power required to process it.

As the technology advances, new product possibilities emerge, such as autonomous vehicles, adding new challenges. “The location technology consumption model has moved online, so we need reliable, scalable MLA technology that won’t let us down,” says Denecker. “Plus, the total volume of data and compute processing is going up tremendously.” Denecker estimates that his teams process a million times more data at the source for every kilobyte of information delivered to a map.

Historically, that processing took place in a TomTom on-premises datacenter, using extensive graphics processing unit (GPU) computing with NVIDIA GPUs. Extracting a single feature from the world map could take six months—a model that didn’t deliver the speed TomTom needs for the innovation edge that keeps it competitive. Denecker’s team found the requisite speed and agility in the cloud.

Taking mapmaking to the cloud

TomTom’s choice of a cloud provider hinged on more than performance metrics. The company’s leaders are acutely aware that attracting first-rate engineering talent is key to TomTom’s innovation goals. “If we can’t offer engineers a world-class, end-to-end developer experience in-house, we’ll struggle to bring them onboard,” says Denecker. “We needed a cloud environment that’s not only cost-effective but also covers the breadth of our needs—performance, compliance, safety, and privacy—in a full stack. That’s why we chose Azure.”

That choice illustrates the trust between TomTom and Microsoft. “Our comprehensive solutions take years to build, and they’ll run for years,” adds Denecker. “Selecting Azure was a significant strategic commitment for TomTom.”

TomTom trusts Microsoft to deliver the performance it needs to satisfy customers’ demand for instantly available mapping information. And it trusts Azure solutions to support confidentiality. “The combination of security, safety, and privacy is a strong requirement in the automotive industry,” explains Martijn Siereveld, Head of the TomTom Cloud Center of Excellence (CCoE). “We’ve got to balance those needs on one hand with at-scale, near real-time map data processing on the other to meet and exceed our customer expectations. We’re only as good as the foundation we build upon, and our relationship with the Azure team supports that foundation.”

The company’s diverse product base exacts different requirements on its infrastructure. The mapping application is one example of TomTom’s need for scalable, responsive storage and computing power. Its Probe technology, which ingests information from vehicles and mobile devices to provide minute-to-minute updates on road conditions and hazards, naturally complements those digital maps, superimposing that dynamic data over the more static topographical images. Prior to its Azure migration, the company processed its Probe data on-premises on Hadoop, an open-source framework for storing and processing enormous datasets. Even with 100 Hadoop nodes, performance couldn’t keep up with demand.

Delivering lightning-fast performance with Azure

The Probe technology bombards the TomTom back end with data—more than a million events per second. As TomTom migrated applications to Azure, it optimized costs by making full use of the Premium tier of Azure Blob Storage and the Hot, Cool, and Archive storage options. The tier’s high input/output operations per second (IOPS) capacity, which is the number of requests a user’s application sends to storage disks in one second, is key to reducing latency for TomTom applications. “We need a high-performance back end just to ingest events,” says Siereveld. “We then use that data for modeling to provide visualization, to experiment, and to build other solutions, which involves very high and on-demand IOPS requirements.” Using Blob Storage also reduced computing power needs dramatically, further creating significant savings.

Developers throughout TomTom use the Azure Storage technologies that best fit their needs at the most advantageous cost, including Azure Databricks clusters, Azure Data Lake Storage, and Azure HDInsight. The team found that the Premium tier of Blob Storage delivered maximum speed, stability, and cost-effectiveness for many of its evolving needs. TomTom developers began using the solution in production in 2019. They appreciate its reliability and agility—they can build and operate solutions without having to involve an infrastructure team.

Developer teams that are working on other applications also use Azure Spot Virtual Machines to optimize costs, pulling data into unused storage capacity for significant cost savings. TomTom underpins its modernization strategy with Azure Kubernetes Service (AKS), running map APIs on the service. “AKS is more than a reliable runtime for dynamic workloads,” says Denecker. “Using it helps us achieve reliable deployment.”

The TomTom Map Operations team built its archive and entire engineering solution on Azure infrastructure, confident the scalability would match the fluctuating demand that characterizes mapping applications like Probe. “As traffic on the roads picks up and load increases, so does performance,” says Denecker. “Not only does Map Operations have solid reliability, but it can expose the data for other engineering teams. It doesn’t have to worry about infrastructure being a chokepoint that constrains sharing data in a highly secure way with another team. And the team was impressed by the ease of use, reliability, and performance, especially the read IOPS it was getting from the back end.”

Sparking innovation for lasting gains

TomTom plans to continue zealously expanding its market share, innovating to stay ahead of demand and set industry trends. “As we process ever more data for new use cases, we need to innovate fast,” declares Denecker. “Our TomTom engineering community must be unconstrained by infrastructure while they innovate, scale up new products, and optimize the delivery of location technology products to our customers around the world.”

Denecker’s goal is for the company to run at the same level of reliability and speed as the cloud. “I think we’re living in a cloud-centered universe,” he concludes. “In the race to be the best map company, our service platform team is the tarmac, supported by our close relationship with Azure.”

Find out more about TomTom on Twitter, Facebook, and LinkedIn.