[
https://issues.apache.org/jira/browse/SPARK-22683?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16355920#comment-16355920
]
Thomas Graves commented on SPARK-22683:
---------------------------------------
If the config is set to 1 which keeps the current behavior the job server
pattern and really any other application by default won't be affected. I don't
see this as any different then me tuning max executors for example. Really
this is just a more dynamic max executors.
I agree with you that this isn't optimal in ways. For instances it applies it
across the entire application where you could run multiple jobs and stages.
Each of those might not want this config, but that is a different problem where
we would need to support per stage configuration for example. If its a single
application then you should be able to set this between jobs programmatically
if they are serial jobs (although I haven't tested this), but if that doesn't
work all the dynamic allocation configs would have the same issue.
> DynamicAllocation wastes resources by allocating containers that will barely
> be used
> ------------------------------------------------------------------------------------
>
> Key: SPARK-22683
> URL: https://issues.apache.org/jira/browse/SPARK-22683
> Project: Spark
> Issue Type: Improvement
> Components: Spark Core
> Affects Versions: 2.1.0, 2.2.0
> Reporter: Julien Cuquemelle
> Priority: Major
> Labels: pull-request-available
>
> While migrating a series of jobs from MR to Spark using dynamicAllocation,
> I've noticed almost a doubling (+114% exactly) of resource consumption of
> Spark w.r.t MR, for a wall clock time gain of 43%
> About the context:
> - resource usage stands for vcore-hours allocation for the whole job, as seen
> by YARN
> - I'm talking about a series of jobs because we provide our users with a way
> to define experiments (via UI / DSL) that automatically get translated to
> Spark / MR jobs and submitted on the cluster
> - we submit around 500 of such jobs each day
> - these jobs are usually one shot, and the amount of processing can vary a
> lot between jobs, and as such finding an efficient number of executors for
> each job is difficult to get right, which is the reason I took the path of
> dynamic allocation.
> - Some of the tests have been scheduled on an idle queue, some on a full
> queue.
> - experiments have been conducted with spark.executor-cores = 5 and 10, only
> results for 5 cores have been reported because efficiency was overall better
> than with 10 cores
> - the figures I give are averaged over a representative sample of those jobs
> (about 600 jobs) ranging from tens to thousands splits in the data
> partitioning and between 400 to 9000 seconds of wall clock time.
> - executor idle timeout is set to 30s;
>
> Definition:
> - let's say an executor has spark.executor.cores / spark.task.cpus taskSlots,
> which represent the max number of tasks an executor will process in parallel.
> - the current behaviour of the dynamic allocation is to allocate enough
> containers to have one taskSlot per task, which minimizes latency, but wastes
> resources when tasks are small regarding executor allocation and idling
> overhead.
> The results using the proposal (described below) over the job sample (600
> jobs):
> - by using 2 tasks per taskSlot, we get a 5% (against -114%) reduction in
> resource usage, for a 37% (against 43%) reduction in wall clock time for
> Spark w.r.t MR
> - by trying to minimize the average resource consumption, I ended up with 6
> tasks per core, with a 30% resource usage reduction, for a similar wall clock
> time w.r.t. MR
> What did I try to solve the issue with existing parameters (summing up a few
> points mentioned in the comments) ?
> - change dynamicAllocation.maxExecutors: this would need to be adapted for
> each job (tens to thousands splits can occur), and essentially remove the
> interest of using the dynamic allocation.
> - use dynamicAllocation.backlogTimeout:
> - setting this parameter right to avoid creating unused executors is very
> dependant on wall clock time. One basically needs to solve the exponential
> ramp up for the target time. So this is not an option for my use case where I
> don't want a per-job tuning.
> - I've still done a series of experiments, details in the comments.
> Result is that after manual tuning, the best I could get was a similar
> resource consumption at the expense of 20% more wall clock time, or a similar
> wall clock time at the expense of 60% more resource consumption than what I
> got using my proposal @ 6 tasks per slot (this value being optimized over a
> much larger range of jobs as already stated)
> - as mentioned in another comment, tampering with the exponential ramp up
> might yield task imbalance and such old executors could become contention
> points for other exes trying to remotely access blocks in the old exes (not
> witnessed in the jobs I'm talking about, but we did see this behavior in
> other jobs)
> Proposal:
> Simply add a tasksPerExecutorSlot parameter, which makes it possible to
> specify how many tasks a single taskSlot should ideally execute to mitigate
> the overhead of executor allocation.
> PR: https://github.com/apache/spark/pull/19881
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
