This is the first part of the research that has a paper published in EDB 2017 conference. The second part descibes about the extended research has an accepted paper in the Journal of Information Sciences and Engineering (JISE) – an SCIE journal.
Data fragmentation in ﬂash SSDs which is a common problem that leads to performance degradation, especially when the underlying storage devices become aged by heavily updating workloads. This paper addresses that problem in MongoDB, a popular document storage in the current market, by introducing a novel stream mapping scheme that based on unique characteristics of MongoDB. Empirical results showed that in YCSB and Linkbench, our methods improved the throughput by more than 44% and 43.73% respectively; reduced 99th-percentile latency by up to 29% and 24.67% in YCSB and Linkbench respectively. In addition, by tuning the leaf page size in B+Tree of MongoDB, we can signiﬁcantly improve the throughput by 3.37x and 2.14x in YCSB and Linkbench respectively.