Choice of time characteristic and performance

[Bob Tiernay]

I was wondering if the choice of time characteristic (ingestion, processing
or event time) makes a difference to the performance of a job that isn't
using windowing or process functions. For example, in such a job is it
advisable to disable auto wartermarking and use the default? Or is this in
combination an explicit choice of one characteristic more optimal?

More generally it would be good to know how this choice effects a job.

Anyone have any details or empirical evidence about this?



--
Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/

[rmetzger0]

Hi Bob,

if you don't need any time characteristics, go with processing time.

Ingestion time will call System.currentTimeMillis() on every incoming record, which is an somewhat expensive call.

Event time (and ingestion time) will attach a long field to each record, making the records 8 bytes larger and the serialization a bit more involved.

Hope this helps!

On Fri, May 14, 2021 at 11:54 PM Bob Tiernay <[hidden email]> wrote:

I was wondering if the choice of time characteristic (ingestion, processing
or event time) makes a difference to the performance of a job that isn't
using windowing or process functions. For example, in such a job is it
advisable to disable auto wartermarking and use the default? Or is this in
combination an explicit choice of one characteristic more optimal?

More generally it would be good to know how this choice effects a job.

Anyone have any details or empirical evidence about this?



--
Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/

[Bob Tiernay]

Thanks for you guidance Robert!

Do you think disabling watermarking would help in terms of a slight
performance boost in such scenarios?

Bob



--
Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/