Abstract:

Meteorological archival and retrieval system (MARS) enables archival and retrieval management for meteorological data,  including many numerical prediction products. However, with the refinement of numerical forecasting techniques, the data under the MARS management is soaring in a massive growing trend, which presents a new challenge to archival and retrieval techniques. As an archival issue in the physical view of MARS, the metadata index currently has very low update efficiency. To address the problem, we propose a parallel approach to accelerate index update speed, which reduces the index reorganizing overhead during data archiving time. Experimental results show that compared with the original serial one, our parallel approach accelerates the maximum update speed by 3 times on average, hence improving the efficiency of meteorological data archiving.

Key words: meterological archival and retrieval system (MARS);physical index structure;parallel updating;rapid archiving