Downloads

Software

XXL: eXtensible and fleXible Library
(Java-library for advanced query processing)
JEPC: Java Event Processing Connectivity
(Java-middleware for uniform event processing)
Data Structure Navigator
(Visualization of data structures and algorithms)
DataGenerator of Norbert Beckmann
R*-Tree-Implementations (e.g. Revised R*-tree, R*-tree, Hilbert R-tree, …) of Norbert Beckmann

This page serves as a source for datasets which we have used in our experiments.

Spatial Data

No.	Area	Description	number of MBRs	zipped size in MB	coverage	source	used for experiments in
M1	L.A.	streets	131,461	1.35	0.03	Tiger	[BKS 93], [BSS 00], [DS 00]
M2	L.A.	rivers and railways	128,971	1.99	0.22	Tiger	[BKS 93], [BSS 00], [DS 00]
M3	california	streets	1,888,012	16.93	0.12	Tiger	[BSS 00], [DS 00]
M4	california	railways	625,640	0.33	0.21	Tiger
M5	california	borders	234,251	2.82		Tiger
M6	california	hydrography	360,330	4.12		Tiger

The temporal data we used in our experiments with the multi-version B-tree is here (small size 100'000).

Datasets for MVBT experiments

Description	format description
d50	ASCII file contains sequence of 10'000'000 triples (blank separated). Triples have following format: Operation type, long key, integer paylod. 1 decodes insert, 2 decodes update and 3 decodes delete; The first 1'000'000 operations are insertions (10% of the data set). The remaining 90% of the file represent a mix of insertions, deletions and updates. The portion of the specific operation is decoded in the file name. For example the file d50 consists of 1'000'000 insert operations followed by a mix of insertions ($4'500'000$) and deletions (4'500'000). The file u75 consists of 1'000'000 insert operations followed by a mix of insertions (2'250'000) and updates 6'750'000. Note default payload value is 0. Please, replace it for your purpose. In our experiments we replace it with 16 bytes paylod. Version numbers generated while reading the file line by line.
u0
u25
u50
u75
u100

Results of Spatial Joins

Description	number of results	format description	zipped size in MB
M2 & M1	85,854	M2.ID M1.ID (the MBRs of M2 are numerated starting with 10,000,000, the MBRs of M1 with 0)	0.37
M3 & M3	9,784,072

Results of k-nearest neighbors

Description	format description	size in MB
20-nearest neighbors for each element in M2	M2.ID M1.ID k 'euclidian distance' (the center of the MBRs was used for the computation, the MBRs of M2 are numerated starting with 10,000,000, the MBRs of M1 with 0)	33.79

Datasets for Sort-based Parallel R-tree loading ([ASSS 2012])

Description	format description	size in MB
USA-data	Contains the minimum bounding rectangles of all streets from TIGER files, containing 72 Million rectangles. The file is in hadoop sequence file format with datasets < NullWritable, DoublePointRectangle>. For convenience, we provide a plain data set consisting of rectangles with the following format: <xlow, ylow, xhigh, yhigh>, each coordinate occupying 8 bytes in double floating point format. The file can be obtained from here.	3338
E-USA-data	Extended USA dataset, composed of four copies of USA-data by translating the original data set with the following vectors: (0.0, 0.0), (75.5, -33.9), (0.0, -33.9), (75.5, -3.9). A plain file can be obtained from here (see USA-data for the file formats).	13414
qr1	Query point data set obtained by considering every 100-th middle point of the rectangles from USA-data, consisting of 722,261 points. The files are in plain format, sequential point data with <x,y> coordinates, each coordinate occupying 8 bytes in double floating point format.	22
qr2	Query rectangle data with quadratic rectangles where each rectangle returns 100 results on average, consisting of 722,226 rectangles. The file format ist the same as for qr1.	2.2
qr3	Query rectangle data with quadratic rectangles where each rectangle returns 1000 results on average, consisting of 22,856 rectangles. The file format ist the same as for qr1.	0.7