We have hit some issues where customers are using smaller heaps, per the recommendations in the Neo4j Operations Manual, since that guidance is for subgraph, OLTP type queries.
With Graph Algorithms, many of them are touching and/or updating large parts of the graph, or the whole graph, and as such need much much larger heaps.
We may even need some info in there pointing to OLAP design considerations, like dedicated Read Replicas for algorithms to run against, which would have larger heaps, and whatever other recommended server level configs for large graph operations, so as to keep the Core member OLTP tuned.
From https://trello.com/c/h1ebBz2J/1736-graph-algos-add-section-to-graph-algorithms-docs-describing-heap-memory-considerations-for-algorithms-olap-vs-the-traditional-ne
We have hit some issues where customers are using smaller heaps, per the recommendations in the Neo4j Operations Manual, since that guidance is for subgraph, OLTP type queries.
With Graph Algorithms, many of them are touching and/or updating large parts of the graph, or the whole graph, and as such need much much larger heaps.
We may even need some info in there pointing to OLAP design considerations, like dedicated Read Replicas for algorithms to run against, which would have larger heaps, and whatever other recommended server level configs for large graph operations, so as to keep the Core member OLTP tuned.