best practice approach for grouping time series data

[Patrick Wiener <pa...@web.de>]

Hello everybody,

I ran into a problem regarding GlobalGrouping time-sensitive data. KafkaSpout emits tuples where one tuple-field contains a timestamp.
Now, I do some processing on the data. In the end my aim is to combine the processed tuples with a GlobalGrouping approach and save it to Redis. From there it will be fetched via Pub/Sub and displayed on a simple Webinterface (its mandatory that the data stream exiting Storm is in the same order as when it entered).

Question:
Is there some kind of best practice approach for the bolt which collects all the tuple streams ? In other words: can the bolt hold back a tuple in order to achieve the right order (tuple n always gets processed/stored before tuple n+1 and so on)?

Patrick