Hello @LSIMS ,

Based on my understanding you are seeing slow reads in general including the count(*) from Oracle itself.. based on that I'd suggest the following:

- If fetch(read) is slow that means the upstream data is not optimized, a good optimization in this case is to index your oracle data and use the index column in your spark JDBC read query.

- I understand you mentioned 10 parallel, 10 parallel will mean 10 workers' cores can run at a time otherwise if your cluster has less cores available the 10 parallel runs will not happen at the same time, in this case please check SparkUI to confirm how many executors are spun and how many cores are used