Hello!
I would like to use Esper in a Storm Bolt. In my use case I have to
scale out the number of machines that are processing the incoming data
(sensors data).
The basic data model sent by sensor is : {sensorId, value, geoLocation}
In order to keep data consistent in the Esper Bolt, the Esper Bolt will
consume from the Spout
grouped by the sensorId.
When starting a new machine, we should rebalance our topology in order
that new machine to process data. In this case, I' afraid that the Esper
Bolt will loose the context that it worked and will not process some
tuples from the spout.
Therefore, my questions are:
1. Is an EsperBolt suitable for scaling up applications?
2. Is it possible with Storm to integrate Esper in such a scenario
(remember, we should assure that we have processed all the tuples before
rebalancing)?
I look forward for your answers and suggestions.
Thank you.
Regards,
Florin
