Amazing that you’ve managed to test and everything works as expected
We don’t have a public benchmark comparing Neo and Memgraph at this point. But we have this Memgraph 1.1 Up to 50% Better Memory Usage and Higher Throughput. Benchmarks are very complex in general, and it’s hard to make something representative for most people. Neo is pretty fast with the cache from our observations, but things became tricky when the cache is too small or in the concurrent environment. Memgraph is good in both cases because 1) Even though the graph is stored in memory, we heavily optimize memory usage. With graphs, a lot of analytics run globally, and the whole graph at some point has to be loaded to the main memory. 2) At scale and databases, concurrency is a big thing because of MVCC, in-memory, and optimized data structures, Memgraph is dealing with concurrent graph algorithms well.
But yes, in simple benchmarks, both of these preconditions are relatively hard to setup + you might not need that initially.
If you could share the most representative queries, that would be amazing? Memgraph is still a very early product. I’m pretty sure we can improve almost any workload.