Use of streaming K Means

[Marko Dinić]

Hello again,

I have again a question about Streaming K Means, concretely about its 
use. From the paperworks that are given I understood that one of its 
advantages is that does one pass clustering and in that way can decrease 
number of iterations and work with large datasets.

What I'm interested in is - can I use it in online fashion? If I have 
data streaming from some data, can I use it to cluster incoming data in 
some way?

I understand that there is streaming step, that in some way looks more 
or less appropriate for my incoming data, but then there is a Ball K 
Means step, that is performed after streaming step. The question that 
arrives is - when to do Ball K Means step, since the data arrives all 
the time...

Should I even consider this, or should I go for lambda architecture?

Any help would be great.

Thanks,
Marko