Large Graph Datasets

This record contains five large graph datasets: orkut, pokec, livejournal, ldbc, and friendster. The ldbc dataset is generated by the LDBC benchmark with scale factor 10. The rest are originally from SNAP. These datasets use a CSR (Compressed Sparse Row) binary format, structured as follows: The file begins with two 32-bit integers representing the vertex count $|V(G)|$ and edge count $|E(G)|$. Next comes the offset array, consisting of $|V(G)| + 1$ 64-bit unsigned integers (unsigned long long), where the $i$-th entry points to the start of vertex $i$'s edges in the edge array (0-based), and the last entry equals $2 \times |E(G)|$. Following this is the vertex label array, storing $|V(G)|$ 32-bit integers, with the $i$-th value representing the label of vertex $i$. Finally, the edge data consists of $2 \times |E(G)|$ 32-bit integers, organized as consecutive destination vertices for each edge, with all edges sorted by their source vertex as defined by the offset array.