Hi Stephan,
Thanks a lot for your response. I’ll study the options that you mention, I’m not sure if the “chagelog stream” will be easy to implement
since the lookup is based on matching IP ranges and not just keys.
Regards,
Diego
De: Stephan Ewen [mailto:[hidden email]]
Enviado el: lunes, 30 de enero de 2017 17:39
Para: [hidden email]
Asunto: Re: Calling external services/databases from DataStream API
Hi!
The Distributed cache would actually indeed be nice to add to the DataStream API. Since the runtime parts for that are all in place, the code would be mainly on the "client" side that sets up the JobGraph to be submitted and executed.
For the problem of scaling this, there are two solutions that I can see:
(1) Simpler: Use the new asynchronous I/O operator to talk with the external database in an asynchronous fashion (that should help to get higher throughput) https://ci.apache.org/projects/flink/flink-docs-release-1.2/dev/stream/asyncio.html
(2) More elaborate: Convert the lookup database into a "changelog stream" and make the enrichment operation a "stream-to-stream" join.
Greetings,
Stephan
On Mon, Jan 30, 2017 at 1:36 PM, Jonas <[hidden email]> wrote:
I have a similar usecase where I (for the purposes of this discussion) have a
GeoIP Database that is not fully available from the start but will
eventually be "full". The GeoIP tuples are coming in one after another.
After ~4M tuples the GeoIP database is complete.
I also need to do the same query.
--
View this message in context: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/Calling-external-services-databases-from-DataStream-API-tp11366p11367.html
Sent from the Apache Flink User Mailing List archive. mailing list archive at Nabble.com.
| Free forum by Nabble | Edit this page |