Challenges of Migration
Uncertainty
Spotify had never embarked on a CPU architecture change, so there was no internal playbook. This was a first for the company.
Complexity
Transitioning from a single architecture x86 environment to a hybrid, and eventually predominantly Arm-based infrastructure, added some complexity in deployment, testing, and reliability.
Resource Prioritization
As with any major change initiative, the Platform team faced prioritization decisions—weighing the impact of allocating engineering resources to the migration. But Spotify was convinced that the migration was the right move.
How Axion Delivered for Spotify
Google Axion, based on Arm Neoverse V2 cores, provides leading price-performance across cloud-based instances. It also integrates with the C4A VM family underpinned by Titanium offload technology, offering 72 vCPUs, DDR5 memory, and Tier 1 networking.
The custom silicon delivers up to 65% better price-performance than comparable x86 VMs on Google Cloud, up to 60% better energy efficiency and performance per watt, and broad support for general-purpose workloads, including web services, databases, and AI inference. Axion aligns with Spotify on key priorities: cost, performance, and sustainability.
Spotify’s Cloud infrastructure migration to Axion involves multiple services and workloads, including Java applications, perimeter services, caching layers, data analytics, ML pipelines, CI/CD, build systems, and more. With most developer client devices also running on Arm-based hardware, Spotify benefits from optimal performance across their organization—from cloud to edge.
Migration Best Practices From Spotify
- Experiment: Go wide when testing which workloads would be easy or hard to migrate, as they will not behave in the same way.
- Evaluate instance families regularly: Familiarity with older instances can create engineering inertia, but using newer, modern instances like C4A brings performance benefits.
- Engage with your cloud service provider: Spotify engaged early with Google to troubleshoot issues and optimize workloads. The Google team helped with small configuration changes, such as upgrading to the latest open-source libraries and enabling Arm-specific build flags that helped fix perceived bottlenecks.
Arm Cloud Migration is our developer initiative for companies of all sizes migrating to Arm architecture in the cloud. Explore the Arm Software Ecosystem Dashboard, network with the Arm community, or access Engineering Experts for migration support. Access additional developer resources from the collaboration between Arm and Google Cloud.