Comparison of LSM indexing techniques for storing spatial data

Journal of Big Data

Table 1 Sorted run sizes of the four compared merge policies, where newer sorted runs are on the left

Merge policy	20 Flushes	40 Flushes	60 Flushes	80 Flushes	100 Flushes	120 Flushes
Binomial (\(k=4\))	1, 4, 15	2, 3, 20, 15	10, 50	10, 20, 50	15, 35, 50	1, 3, 10, 106
Tiered (\(B=4\))	4, 16	4, 4, 16, 16	4, 4, 4, 16, 16, 16	16, 64	4, 16, 16, 64	4, 4, 16, 16, 16, 64
Concurrent (default)	3, 17	1, 1, 3, 35	1, 59	3, 77	1, 1, 3, 95	1, 119
Leveled (\(B_0=2, B=4\))	1, 1, 4, 14	1, 1, 4, 16, 18	1, 1, 4, 16, 38	1, 1, 4, 16, 58	1, 1, 4, 16, 64, 14	1, 1, 4, 16, 64, 34

Each number is the size of a sorted run with respect to the MemTable size. Tiered, Concurrent and Leveled always have sorted runs in non-decreasing order. The first two sorted runs in Leveled policy are two disk components in level 0, the other numbers are the number of disk components of size 1 in the corresponding levels. Default parameters for Concurrent: \(k=30, C=3,D=10, \lambda =1.2\).