venkata91 opened a new issue, #17563:
URL: https://github.com/apache/iceberg/issues/17563

   ### Feature Request / Improvement
   
   Iceberg publishes scan bytes and file/manifest counts per task, but the only 
timer in the metrics system is `totalPlanningDuration`, which is started in 
`SnapshotScan` around driver-side planning. `CommitMetrics.totalDuration` 
covers commit, not data movement.
   
   The result is that scan throughput cannot be derived from Iceberg metrics 
alone. Bytes are available, elapsed read time is not, and task duration is a 
poor proxy because it also includes everything the query does after the scan 
(in one workload we looked at, shuffle write was ~94% of task runtime).
   
   This matters when the question is about the read path itself, for example 
comparing the same table read from two clusters to see whether one is crossing 
regions, or telling a slow scan apart from a slow downstream operator.
   
   For reference, Spark exposes `scanTime` on `FileSourceScanExec`, but it is 
defined only for that operator, so a DSv2 `BatchScan` gets nothing. 
`lance-spark` publishes `batchLoadTimeNs` for the equivalent read path.
   
   Proposal: add a per-task scan-duration metric alongside the existing 
counters, aggregated to the stage in the same way as `totalPlanningDuration`. 
Nanoseconds rather than milliseconds, since a single split can be read in well 
under a millisecond.
   
   PR: https://github.com/apache/iceberg/pull/17562
   
   ### Query engine
   
   Spark
   
   ### Willingness to contribute
   
   - [x] I can contribute this improvement/feature independently
   - [ ] I would be willing to contribute this improvement/feature with 
guidance from the Iceberg community
   - [ ] I cannot contribute this improvement/feature at this time


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to