Wail,
The fact that micro-operators are there and are only in Algebricks was
because at the early stage of Hyracks, data transferred between operators
went through network interfaces regardless they are in the same machine or
not and a pipeline of micro-operators can stay in the same full operator
and hence can avoid extra memory copies.
However, a long time ago, we added job-levels rewritings in Hyracks to
pact operator pipelines (and DAGs) that are local into the same "bigger"
operator (SuperActivity) to achieve the same goal.
So, long story short, you don't need to care the micro operators
anymore and Hyracks will do the optimization (rewriting) automatically for
full operators:-)
Best,
Yingyi
On Fri, Aug 28, 2015 at 12:06 PM, Wail Alkowaileet <[email protected]>
wrote:
> Hi all,
>
> I want to ask what are the differences between the micro and full Hyracks
> operators?
> What kind of operations that are suitable for each one? And does the Role
> Optimizer can convert from one to another or it's the responsibility of the
> developer ?
>
> Thanks!
> --
>
> *Regards,*
> Wail Alkowaileet
>
