答复: About the balance of data under field grouping

[Junfeng Chen <k-...@hotmail.com>]

That should depend on how you want to process your tuples. 

According to the Storm document, fieldgrouping can be used in the situation
that the tuples with same field value must be processed in the same task,
like Word Count sample. However I have no idea about your purpose. 

As in your scenario the load of one task may be much heavier than others, it
is not recommended to use  field grouping in your program. 

 

发件人: 姚驰 [mailto:yaochitc@163.com] 
发送时间: 2014年11月3日 14:21
收件人: user@storm.apache.org
主题: About the balance of data under field grouping

 

Hi all,

  As I'm not so familiar with storm's patterns, I have a question about
field grouping. I found that the tuple with same field value

with be sent to the task(the same thread) under fieldgrouping, so what
should I do if I have a stream in which some field values have far more

tuples than the others? (I mean, the amount of tuples processed by each
thread will be quite different if I want to  do something like counting or
computing with fieldgrouping.) Should I use fieldgrouping to solve it or use
others patterns?