It looks like this sharded representation allows fast access for pretty much any large graph and pretty much any computation we want to perform on the graph, because the data is very well organized if you only need to process the graph in chunked subgraphs, but it also allows fairly quick access to any of the edges/neighbors for a given vertex because we store vertices by subgraph and ordered by source.
It looks like this sharded representation allows fast access for pretty much any large graph and pretty much any computation we want to perform on the graph, because the data is very well organized if you only need to process the graph in chunked subgraphs, but it also allows fairly quick access to any of the edges/neighbors for a given vertex because we store vertices by subgraph and ordered by source.