Hi, Thank you for your reply.
I would like to cluster the values (every message contains one value) after their region and to add the values up in realtime afterwards (the actual/last minute). I would like to use Kafka to subscribe the messages to a stream processing framework like Storm or Spark Streaming and to use a time window because I would like to add up the values of the last minute. Would it be better to use different topics for different regions or to use one topic and to do the clustering in Storm or Spark Streaming afterwards? Thank you in advance. Regards, Daniela Gesendet: Mittwoch, 23. März 2016 um 09:42 Uhr Von: "Manikumar Reddy" <ku...@nmsworks.co.in> An: "users@kafka.apache.org" <users@kafka.apache.org> Betreff: Re: Topics in Kafka Hi, 1. Based on your design, it can be one or more topics. You can design one topic per region or one topic for all region devices. 2. Yes, you need to listen to web socket messages and write to kafka server using kafka producer. In your use case, you can also send messages using Kafka Rest Proxy from confluent. http://docs.confluent.io/2.0.1/kafka-rest/docs/index.html On Wed, Mar 23, 2016 at 1:59 PM, Maria Musterfrau <daniela_4...@gmx.at> wrote: > Hi > > I am new and I have a question regarding Kafka. I have devices in > different regions. Every device opens a websocket connection when it gets > active and it sends its messages over the opened websocket connection to a > server. My question is: is every region a topic or is every websocket > connection (means every device) a topic? Or can I choose which one is used > as topic? > > Is it true that I need a piece of code as Kafka cannot consume messages > directly from the websocket connection? > > Thank you and regards, > Daniela >