ASML: Accelerating Microchip Testing with Google Cloud

ASML is the world’s largest and most advanced supplier of photolithography machines, which are essential for the manufacture of microchips. Based in Veldhoven, Netherlands, the company controls more than 60% of the global market for photolithography machines and is the only producer of the most-accurate EUV (extreme ultraviolet) systems. These systems enable the creation of next-generation, hyper-miniaturized chips. Each EUV machine is the size of a double-decker bus, weighs 180 tons, and sells for more than €100 million. ASML's machines play a vital role in the microchip industry, and their software needs to be ready to meet any challenge due to the high cost of production delays and the potential for microscopic faults.

ASML, the world’s leading producer of photolithography machines essential for microchip manufacturing, was facing a significant challenge in its testing process. The company's rigorous testing procedures, which are crucial due to the high cost of production delays and the potential for microscopic faults, were time-consuming. Engineers were provided with virtual machines for testing, run from an on-premises data center, but these could take a full day to return results. This delay was holding back the development cycle, as software engineers had to wait anywhere from two hours to a full day for test results, due to limited on-premises data center capacity. The demand for testing varied dramatically, depending on the number and scale of tests, making it difficult to manage with the existing infrastructure.

To address this challenge, ASML, with the help of cloud consultancy Rackspace, migrated to a cloud-based solution using Google Cloud. They drafted a microservices architecture orchestrated by Google Kubernetes Engine. However, they decided to go a step further and create an ultra-lean, serverless architecture using Google Cloud products. This eliminated the need to maintain a Kubernetes cluster and helped keep ASML within project usage quotas on Google Cloud. ASML uses Google Cloud Functions and Google Cloud Build to deliver test VMs for its engineers without the need for dedicated infrastructure. The system works by sending a test request and its payload to a Google Cloud Storage bucket, which triggers a Google Cloud Pub/Sub message associated with a Google Cloud Function. This function evaluates the needs of the request and calls a Cloud Firestore database to specify the VMs required. A Cloud Function then calls a Cloud Build, which creates that infrastructure with the Google Compute Engine API and begins to run tests with content from the bucket.

• The new cloud-based solution has significantly improved the testing process at ASML. It has not only reduced the time taken to test code by around 80%, but also eliminated the need for queueing, making the process more predictable. The immediate availability of cloud infrastructure allows ASML to scale at speed to match demand without the need to provision excess infrastructure and pay for maintenance. This has driven a culture of rapid testing, encouraging innovation and rooting out issues more effectively. The solution is already being used by more than 300 engineers, and certain tests, such as the full qualification of the code, which currently runs daily, will soon run every time a line of code is changed.

• Delivers test results in under one hour, instead of two hours to a day

• Operates a serverless solution without the need to run or maintain a cluster

• Achieves 80% reduction in time taken to test code