Trying to implement Getafix system for optimizing segment replication factor and Query Latency

Hi all,

I read this Paper -

https://www.ideals.illinois.edu/bitstream/handle/2142/89289/getafix.pdf?sequence=4

It implements a system called GetaFix which analyses the segment popularity by emitting metrics from Historicals to the coordinator , which runs an algorithm for finding out the Best Replication Factor of segments in the historical in order to optimize the storage factor in historicals and also minimize the Query Latency . It seems really promising .

What i am confused about is -

  1. Is it possible to be implemented using the current druid APIs ?

If Yes - Then how can i make historicals and coordinators talk for emitting segment popularity metrics and coordinators telling historicals to drop or load certain segments and make brokers and coordinators talk for sending proper routing information of queries.

  1. Do i need to make any changes in the current druid code?

Please give me some insight so that i can move forward in implement this idea.


Mailtrack

Sender notified by

                [Mailtrack](https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality5&)
                22/11/19, 14:51:06