Should i use Trident to compute the global mean of tuples in Storm? -
i want compute storm mean incoming tuples made of [int id,int value]. can see can't partition data using fields grouping. need topology architecture distribute computation , way im thinking of doing mini batches within each bolt instances , aggregate.
i kind of understood trident appropriate solution mini-batch processing within storm.
what best practice compute global analytics storm means, global count, std-devs when can't partition data based on attribute? topology example?
you can compute stream statistics such mean, standard deviation , count computed using trident-ml. there's a section in readme explains how compute theses stats within trident topology.
hope helps.
Comments
Post a Comment