Hi, For any kin of performance analysis, first determine the following 1. How many topologies? 2. What are the workload parameters (e.g. tuple arrival rate, number of executors per component, size of each tuple) 3. System environment ( Number of cores in each supervisor node, Available RAM) 4. What is the nature of the algorithm? Is it CPU intensive?
Next determine the performance metric. Usually, for storm it can be average tuple processing latency/ throughput etc. Then you need to follow the factor at a time approach where you change one workload parameter per experiment and measure the performance metric while keeping others same. This will help you to identify the effect of each parameter on the performance. On Wed, Nov 16, 2016 at 4:46 PM, Matt Lowe <[email protected]> wrote: > Performance in what way? > > If you mean from the storm Twitter example, that is part of the business > logic. It is a way of filtering. > > Best Regards > Matthew Lowe > > On 16 Nov 2016, at 22:34, sam mohel <[email protected]> wrote: > > Please .Is there any help ? > > On Wednesday, November 16, 2016, sam mohel <[email protected]> wrote: > >> How can algoritm like" text similarity between tweets" affect on the >> performance of storm topology ? > > --
