On one of Linagora owned plateform, reindexing of 500M emails is conducted at a speed of 11K email / hour.
This means given our dataset 5 years to conduct the reindexing.
We should speed that up.
The reindexing is performed sequentially, without reactor, and we are paying a heavy price with all sorts of latencies, interaction is performed with 1 store at a time (cassandra, elasticSearch, swift are involved), no parallelism is leveraged.
All steps of reindexing should be done in a reactive fashion.
Acceptance criteria: Performance improvment measured