Vespa Product Updates, December 2020

In December 2020, Vespa announced several significant updates aimed at enhancing the platform's capabilities and performance. These updates include optimizations for tensor performance, particularly for sparse and mixed tensors, resulting in dramatic latency reductions, exemplified by a 10x speedup in end-to-end latency and a 150x speedup for mixed tensor operations. Additionally, Vespa introduced the integration of Apache ZooKeeper for better server cluster management and released the Pyvespa Python library to aid in faster prototyping and machine learning experiments. Moreover, Vespa has integrated ONNX Runtime to expand its capacity for evaluating large models without conversion. Vespa, an open-source big data processing and serving engine developed largely by Yahoo engineers, continues to grow thanks to community feedback and contributions, and it is utilized by numerous Yahoo products and the Verizon Media Ad Platform.