MESSI: In-Memory Data Series Indexing

Abstract

Data series similarity search is a core operation for several data series analysis applications across many different domains. However, the state-of-the-art techniques fail to deliver the time performance required for interactive exploration, or analysis of large data series collections. In this work, we propose MESSI, the first data series index designed for in-memory operation on modern hardware. Our index takes advantage of the modern hardware parallelization opportunities (i.e., SIMD instructions, multi-core and multi-socket architectures), in order to accelerate both index construction and similarity search processing times. Moreover, it benefits from a careful design in the setup and coordination of the parallel workers and data structures, so that it maximizes its performance for in-memory operations. Our experiments with synthetic and real datasets demonstrate that overall MESSI is up to 4x faster at index construction, and up to 11x faster at query answering than the state-of-the-art parallel approach. MESSI is the first to answer exact similarity search queries on 100GB datasets in ∼50msec (30-75msec across diverse datasets), which enables real-time, interactive data exploration on very large data series collections.

Matrix profile goes MAD: variable-length motif and discord discovery in data series

“…Data series 1 have gathered the attention of the data management community for more than two decades (Agrawal et al, 1993;Jagadish et al, 1995;Rafiei and Mendelzon, 1998;Chakrabarti et al, 2002;Papadimitriou and Yu, 2006;Camerra et al, 2010;Kashyap and Karras, 2011;Wang et al, 2013b;Camerra et al, 2014;Dallachiesa et al, 2014;Zoumpatianos et al, 2016;Yagoubi et al, 2017;Jensen et al, 2017;Palpanas, 2017;Kondylakis et al, 2018;Peng et al, 2018;Gogolou et al, 2019;Echihabi et al, 2018Echihabi et al, , 2019Yagoubi et al, 2020;Kondylakis et al, 2019;Peng et al, 2020a;Peng et al, 2020b;Palpanas, 2020;Gogolou et al, 2020). They are now one of the most common types of data, present in virtually every scientific and social domain (Palpanas, 2015;Raza et al, 2015;Mirylenka et al, 2016;Keogh, 2011;Palpanas and Beckmann, 2019;Bagnall et al, 2019).…”

Return of the Lernaean Hydra: Experimental Evaluation of Data Series Approximate Similarity Search

“…Similarity Search. A large number of data series similarity search methods has been studied, supporting exact search [7,137,124,81,127,110], approximate search [136,85,10,46,49], or both [32,134,152,33,163,157,88,96,95,122,158,90,123]. In parallel, the research community has also developed exact [23,67,22,26,44,154,57] and approximate [73] similarity search techniques geared towards generic multidimensional vector data 2 .…”

ParIS+: Data Series Indexing on Multi-core Architectures

“…Thus, tradi-tional solutions and systems are inefficient at, or incapable of managing and processing the voluminous sequence collections that already exist in several domains. Finally, we note that, given the evolution of CPU performance, where the processor clock speed is not increasing due to the power wall constraint, efforts for algorithmic speedups now exploit the parallelism opportunities offered by modern hardware [5,10,35,39,47].…”