Breadth-First Search (BFS) Benchmarks

Lucata powers high performance graph analytics with a next generation computing architecture. Performance benchmarks running Breadth-First Search (BFS) on Graph500 datasets using Lucata Pathfinder are below.

Breadth-First Search (BFS) Benchmarks on Lucata Pathfinder

Database Scale (2^N)	1 Chassis	2 Chassis	4 Chassis
22	215.26	384.69	573.80
24	311.00	529.96	785.42
26	234.08	456.76	852.51
28	247.58	489.32	933.33
29		554.4	1,066.19
30			955.13
* Data in MTEPS (millions of edges traversed per second)

K-hop Benchmarks (k=6) on Lucata Pathfinder

The k-hop graph algorithm uses BFS (breadth-first search) to count all neighbors of a source vertex which are k “hops” away. A source vertex is selected and a count of the number of vertices 6 hops away are counted in the benchmark below.

Database Scale (2^N)	1 Chassis	2 Chassis	4 Chassis
21	0.142	0.176	0.336
22	0.176	0.313	0.548
23	0.227	0.395	0.684
24	0.262	0.562	0.947
25	0.272	0.610	1.056
26	0.309	0.718	1.226
27	0.312	0.592	1.017
28	0.323	0.711	1.185
29		0.889	1.529
30			1.170
* Data in GTEPS (billions of edges traversed per second)