Intra-file-level parallelism on XML parsing: Framework and performance evaluation

Abstract

XML parsing is one of critical operations influencing the performance of XML-based applications. To accelerate its performance on multi-core processor, this paper proposes a novel Intra-File-Level Parallelism (IFLP) framework. Given an XML document, an IFLP parser can split it into small fragments with a little overhead, and simultaneously parse these fragments with several independent processing elements. Finally, the parsing results of the fragments will be combined together. We emulate an IFLP parser based on the well-known Expat parser. Through some experiments and theoretical analysis, we show that in most cases, the IFLP parser can effectively reduce the parsing latency, especially for large documents (i.e. >100KB), the latency can be decreased close to 1/N (N is the number of processing elements).