In the previous chapter, we came across different methods in which RDD could be sampled to give a randomized output. This works fine as long as we want to debug or test our application. However, in other scenarios we might not want to get results which are accurate but which take a long time to execute, but rather require an approximate result within a certain percentage of error and in a time-bound manner. Spark has introduced approximate algorithms to cater to such needs, where the job can guarantee a result within a stipulated timeframe or/and within an error percentage. Some of the actions are:
- countApprox: Returns the approximate value of an RDD within a stipulated time, whether or not the job has completed or not. ...
