Benchmarks

Description of the system

For us to run the experiments we will use a machine proprietary of the Web Semantic Research Group (WESO) of the University of Oviedo. The machine has the following characteristics:

• CPU: Intel(R) Xeon(R) Silver 4214 CPU @ 2.20GHz (12 cores and 24 threads)
• RAM: 40GB
• OS: Ubuntu 20.04.3 LTS

Description of the dataset

We have created the 1-University LUBM dataset, with the following characteristics:

• Subjects:
• Predicates:
• Objects:
• Triples:
• Size: 8.268 MB

RemoteHDT Import (N-Triples to ZARR)

⚠️ Note that the following is just a comparison between the different released versions, and should not be seen as the actual performance of the program in your system.

Version	Description	Threads	Time (s.)	RAM (GB)	Resulting (MB)	Complexity	Compression
v0.0.1	Baseline	1				O(n3)	❌
v0.0.2	Multi-threaded and cartesian product implemented	16				O(n3)	❌
v0.0.3	Loop simplified	1				O(n)	❌
v0.0.4	Compression implemented	1				O(n)	✅
v0.0.5	Multi-threaded, loop simplified and Compression	16	156.12	11.1	4.443558	O(n)	✅
v0.0.6	Multi-threaded, loop simplified and Compression	16	14.08	0.80	204.4581	O(n)	✅

RemoteHDT Deserialization

RemoteHDT vs. SPARQL Queries