About the performance of spark-streaming

i use spark-streaming to write data into memsql, along with the data getting bigger, the performance will geting slower , see the pic below.

when the time process one batch from 4 seconds to more than 50 seconds, there is only 20,000,000 rows,and each row has 13 fields , the total memery usage is 12GB

is there any idea to improve the performance ?

I finally find this issue is my code defect at multithreading, there is some memory leak and I have finally resolved it. In my develop environment, it can reach 200,000 records per seconds at insert or update SQL statement of 9 nodes.

thanks very much.

1 Like