how spark dstream handles congestion?

[Dong Mo]

Dear list,

I was wondering how Spark handles congestion when the upstream is
generating dstreams faster than downstream workers can handle?

Thanks
-Mo