[
https://issues.apache.org/jira/browse/YARN-4039?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14694604#comment-14694604
]
Sadayuki Furuhashi commented on YARN-4039:
------------------------------------------
[~ozawa] Thank you for your review. I attached 2nd patch for trunk code.
I looked into {{weight.getWeight(ResourceType.MEMORY)}} code but I think using
ResourceCalculator there doesn't make sense because value of <weight> in
fair-scheduler.xml is a number. It's not pair of memory and cpu. Therefore
FairScheduler never sets ResourceType.CPU to queue. Only
DominantResourceFairnessPolicy sets and uses ResourceType.CPU without
sideffects to FSQueue, as far as I understand the code.
I ran a simple benchmark. Here is the result:
* cluster configuration:
** total memory: 12GB
** yarn.app.mapreduce.am.resource.mb: 1GB
** mapreduce.map.memory.mb: 1GB
** mapreduce.reduce.memory.mb: 1GB
** maxRunningApps of a queue: 8
** maxAMShare: 0.1
* procedure:
** Issue 12 applications to a queue.
** An application has 10 map tasks and 2 reduce tasks.
*** each map task takes 20 ~ 35 seconds.
*** each reduce task takes 4 ~ 6 seconds.
** Run this benchmark twice and take the shortest duration time.
With demand-blocks-am=false, it took 538.148 seconds to finish all of
applications.
As this screenshot shows, 6 AMs run concurrently and 3 AMs run without running
tasks (Running Containers=1 meaning that only AM is running).
!https://gyazo.com/f82f41b67d2e758cf7af3a5b8991c5dd.png|width=50%!
With demand-blocks-am=true and
yarn.scheduler.fair.continuous-scheduling-sleep-ms=6000, the same benchmark
took took 493.175 seconds.
As this screenshot shows, 2 ~ 3 AMs run at a time.
!https://gyazo.com/ddbe4fe87ba9b738f088c1b209c1a46d.png|width=50%!
!https://gyazo.com/1edd20787e0a17b5b5b8b7920d182895.png|width=50%!
I used this Hive query to run benchmark applications:
{code}
SELECT symbol, COUNT(1) AS cnt FROM nasdaq GROUP BY symbol
union all
SELECT symbol, COUNT(1) AS cnt FROM nasdaq GROUP BY symbol
union all
SELECT symbol, COUNT(1) AS cnt FROM nasdaq GROUP BY symbol
union all
SELECT symbol, COUNT(1) AS cnt FROM nasdaq GROUP BY symbol
union all
SELECT symbol, COUNT(1) AS cnt FROM nasdaq GROUP BY symbol
union all
SELECT symbol, COUNT(1) AS cnt FROM nasdaq GROUP BY symbol
union all
SELECT symbol, COUNT(1) AS cnt FROM nasdaq GROUP BY symbol
union all
SELECT symbol, COUNT(1) AS cnt FROM nasdaq GROUP BY symbol
union all
SELECT symbol, COUNT(1) AS cnt FROM nasdaq GROUP BY symbol
union all
SELECT symbol, COUNT(1) AS cnt FROM nasdaq GROUP BY symbol
{code}
> New AM instances waste resource by waiting only for resource availability
> when all available resources are already used
> -----------------------------------------------------------------------------------------------------------------------
>
> Key: YARN-4039
> URL: https://issues.apache.org/jira/browse/YARN-4039
> Project: Hadoop YARN
> Issue Type: Improvement
> Components: fairscheduler
> Affects Versions: 2.4.0, 2.5.0, 2.6.0, 2.7.0
> Reporter: Sadayuki Furuhashi
> Assignee: Sadayuki Furuhashi
> Attachments: YARN-4039.1.patch, YARN-4039.2.patch
>
>
> Problem:
> In FairScheduler, maxRunningApps doesn't work well if we can't predict size
> of an application in a queue because small maxRunningApps can't use all
> resources if many small applications are issued, while large maxRunningApps
> wastes resources if large applications run.
> Background:
> We're using FairScheduler. In following scenario, AM instances wastes
> resources significantly:
> * A queue has X MB of capacity.
> * An application requests 32 containers where a container requires (X / 32)
> MB of memory
> ** In this case, a single application occupies entire resource of the queue.
> * Many those applications are issued (10 applications)
> * Ideal behavior is that applications run one by one to maximize throughput.
> * However, all applications run simultaneously. As the result, AM instances
> occupy resources and prevent other tasks from starting. At worst case, most
> of resources are occupied by waiting AMs and applications progress very
> slowly.
> A solution is setting maxRunningApps to 1 or 2. However, it doesn't work well
> if following workload exists at the same queue:
> * An application requests 2 containers where a container requires (X / 32) MB
> of memory
> * Many those applications are issued (say, 10 applications)
> * Ideal behavior is that all applications run simultaneously to maximize
> concurrency and throughput.
> * However, number of applications are limited by maxRunningApps. At worst
> case, most of resources are idling.
> This problem happens especially with Hive because we can't estimate size of a
> MapReduce application.
> Solution:
> AM doesn't have to start if there are waiting resource requests because the
> AM can't grant resource requests even if it starts.
> Patch:
> I attached a patch that implements this behavior. But this implementation has
> this trade-off:
> * When AM is registered to FairScheduler, its demand is 0 because even AM
> attempt is not created. Starting this AM doesn't change resource demand of a
> queue. So, if many AMs are issued to a queue at the same time, all AMs will
> be RUNNING. But we want to prevent it.
> * When a AM starts, demand of the AM is only AM attempt. Then AM requires
> more resources. Until AM requires resources, demand of the queue is low. But
> starting AM during this time will start unnecessary AMs.
> * So, this patch doesn't start immediately when AM is registered. Instead, it
> starts AM only every continuous-scheduling-sleep-ms.
> * Setting large continuous-scheduling-sleep-ms will prevent wasting AMs. But
> increases latency.
> Therefore, this patch is enabled only if new option
> "demand-blocks-am-enabled" is true.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
