Details
-
Improvement
-
Status: Closed
-
Major
-
Resolution: Fixed
-
3.2.3
-
None
Description
We currently have a single synchronized iterator by which TinkerGraphComputer vertices are pulled by workers. Moreover, we have a single TinkerMemory by which workers add reductions too. I believe these are two bottleneck concurrency data structures that are causing us performance issues at 1M+ vertices in TinkerGraph (OLAP).