Speaker
Description
High-energy physics analyses repeatedly process the same data as physicists refine algorithms and test new ideas. When the data are read from remote storage, each iteration incurs network latency and depends on shared-storage throughput, which can vary substantially under load. We present uCache (xrd-ucache), a transparent cache implemented as an XRootD client plugin that requires neither server-side deployment nor changes to analysis code. It uses local storage on the analysis machine as a cache layer between the network and memory. The cache stores only the byte ranges read by an analysis and can also rebuild cached data into a branch-aligned, recompressed form, eliminating most of the input/output and decompression costs of subsequent passes. We benchmark uCache using a realistic physics analysis of public CMS data, with both TTree and RNTuple containers, on machines ranging from a 6-core desktop to a 192-core facility node. uCache is open source under the MIT license: https://github.com/drkovalskyi/xrd-ucache