Case Study: Innoviz accelerates AI performance with WEKA file system

Weka Selected by Innoviz to Accelerate AI for Autonomous Vehicle Innovations

Innoviz, a manufacturer of LiDAR sensors and perception software for autonomous vehicles, faced a challenge in accelerating its AI and deep learning workflows. Their legacy NFS-based NAS storage was causing I/O starvation for their GPU cluster, hindering performance and scalability for data-intensive model training. To overcome this, Innoviz selected the Weka File System (Weka) from vendor Weka.

The Weka solution provided a high-performance, scalable file system that delivered extreme bandwidth to Innoviz's GPU servers. This implementation ensured the GPU cluster remained fully saturated with data, significantly reducing AI training time. Weka achieved measurable performance improvements, delivering over 70 GB/s bandwidth to a single node, which was 10x more than their previous NFS solution and 3x more than a local NVMe SSD, enabling faster innovation cycles.

Innoviz

Oren Ben Ibghei

IT Manager