Case Study: Cerence achieves cost-efficient scalability and flexibility with WekaFS from Weka

Cerence™ Gains Superior Cost-Efficiency, Scalability, and Flexibility with WekaFS™

Cerence, a developer of conversational AI for automotive virtual assistants, needed a cost-optimized way to support POSIX-based research workflows while bridging to object storage and the public cloud. As its speech, NLP, NLU, text-to-speech, and predictive keyboard workloads grew, the company sought a scalable, future-proof storage architecture that could reduce datacenter footprint, use NVMe, and handle mounting data volumes more efficiently. WekaFS from Weka was selected as the core platform.

Weka implemented WekaFS on HPE servers in a converged, software-defined architecture that tiers data seamlessly between NVMe flash and object storage, giving Cerence a single namespace with automatic movement of data between on-premises storage and the cloud. The result was a much smaller, more cost-efficient environment with better GPU utilization, simpler administration, and the best performance and lowest latency among the evaluated options. Cerence is managing 900TB on WekaFS and 3.2PB in object storage, with Weka enabling modular scalability and more flexible capacity planning for future growth.

Cerence

Bridget Collins

Chief Information Officer