Hi All, I want to implement adaptive scheduling for our product.
I used custom scheduler. also i go through following links. http://www.dis.uniroma1.it/~midlab/articoli/ABQ13storm.pdf http://www.orgs.ttu.edu/debs2013/presentations/DEBS13-Paper88-Querzoni.pdf I liked the idea of topology based scheduling to take the localization benefit. Any one done this before. Can you please share some code or your experience about this? Text from the above links: --------------------------------- We developed two distinct algorithms based on such idea. One looks at how components are interconnected within the topology to determine what are the executors that should be assigned to the same slot. The other relies on the mon- itoring at runtime of the trac of exchanged tuples among executors. The former is less demanding in terms of required infrastructure and in general produces lower quality sched- ules, while the latter needs to monitor at runtime the cluster in order to provide more precise and e ective solutions, so it entails more overhead at runtime for gathering performance data and carrying out re-schedulings Regards Tarkeshwar
