Case Study: Tencent achieves predictable, high-performance service levels with Alluxio

Tencent News Delivers Exceptional Customer Experience and Predictable Service Levels with Alluxio

Tencent, a major technology company, needed to ensure Tencent News could deliver a fast, reliable experience to more than 100 million active monthly users while meeting strict SLA requirements. Its Spark analytics jobs ran on a dedicated compute cluster but depended on data stored in HDFS outside the cluster, and under peak load the system could not always cache data in memory or complete jobs on time.

To solve this, Tencent deployed Alluxio as an in-memory data layer in the Spark + Alluxio + HDFS stack, separating compute from storage without changing existing applications. Alluxio served data locally from memory after the initial HDFS read, improving performance, enabling predictable job completion, and allowing compute and storage to scale independently; the solution expanded from 150 servers to over 600 nodes.

Tencent