Conference Paper

The Semantic Space-Time Models of the Streamonas Data Stream Management System

Comput. Sci. Dept., Univ. of California Los Angeles, Los Angeles, CA, USA
DOI: 10.1109/CSIE.2009.497 Conference: Computer Science and Information Engineering, 2009 WRI World Congress on, Volume: 4
Source: IEEE Xplore


Four novel models define the fundamental temporal semantics upon which the stream on as DSMS has been architectured, as also the fundamental temporal semantics for application development on the system. Extensive experimental results demonstrate the power of the theoretical models as also the stability and scalability of the system when this is tested with a load from 2, 4, 6, 8 and 10 expressways when running the Linear Road Benchmark. The extensive experimental results demonstrate the effectiveness of the theoretical semantic-space time models as the system has reached the maximum level of difficulty of the benchmark (10 expressways) with an average query latency of 0.000026 seconds, 192,307 times faster than the 5 seconds hard real-time constraint the benchmark sets.

1 Read
  • Source
    • "We wanted to stress test Streamonas in a scenario where patterns are searched within a larger space than the semantic space[2]defined by the 7 stocks. For this reason we performed a second experiment where we applied dynamic clustering of spatio-temporal subsequences over the large semantic space of the speeds of the 1,373,327 cars of the LRB (10 XWays), simultaneously with the dynamic clustering of the 7 stocks (historical span was 6 for all ASTs). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Data Stream Management Systems (DSMSs) have not yet reached a mature enough stage to effectively run data mining algorithms, as they still face challenges within the streaming environment. Streamonas DSMS, as presented in a recent publication, is the first DSMS to reach the maximum level of difficulty supported by the Linear Road Benchmark which is 10 Expressways. The powerful engine of Streamonas can manage an input stream of 20,368 tuples/second with an average query latency of 0.000026 seconds, 192,307 times faster when compared to the 5 seconds maximum query latency the benchmark allows. The on-line data mining over streams presented in this work, is the first effort to apply spatio-temporal data mining algorithms on the Streamonas DSMS system. Dynamic clustering of spatio-temporal subsequences in real-time has been performed successfully, within the large space, high bandwidth, heavy load linear road benchmark streaming platform. Dynamic clustering queries have been expressed in a novel SQL-like language, which we name Streamonas-SQL. Keywords: real-time, data mining, spatio-temporal, dynamic clustering, pattern matching, streamonas, streamonas-SQL, Linear Road Benchmark, query latency, throughput, semantic space.
    Preview · Conference Paper · May 2009