-10.3 C
New York
Monday, December 23, 2024

Analyze Amazon EMR on Amazon EC2 cluster utilization with Amazon Athena and Amazon QuickSight


Gaining granular visibility into application-level prices on Amazon EMR on Amazon Elastic Compute Cloud (Amazon EC2) clusters presents a chance for patrons searching for methods to additional optimize useful resource utilization and implement honest price allocation and chargeback fashions. By breaking down the utilization of particular person functions operating in your EMR cluster, you possibly can unlock a number of advantages:

  • Knowledgeable workload administration – Utility-level price insights empower organizations to prioritize and schedule workloads successfully. Useful resource allocation selections could be made with a greater understanding of price implications, probably bettering total cluster efficiency and cost-efficiency.
  • Value optimization – With granular price attribution, organizations can determine cost-saving alternatives for particular person functions. They’ll right-size underutilized assets or prioritize optimization efforts for functions which are driving excessive utilization and prices.
  • Clear billing – In multi-tenant environments, organizations can implement honest and clear price allocation fashions primarily based on particular person utility useful resource consumption and related prices. This fosters accountability and permits correct chargebacks to tenants.

On this submit, we information you thru deploying a complete resolution in your Amazon Net Companies (AWS) surroundings to investigate Amazon EMR on EC2 cluster utilization. By utilizing this resolution, you’ll acquire a deep understanding of useful resource consumption and related prices of particular person functions operating in your EMR cluster. It will enable you optimize prices, implement honest billing practices, and make knowledgeable selections about workload administration, finally enhancing the general effectivity and cost-effectiveness of your Amazon EMR surroundings. This resolution has been solely examined on Spark workloads operating on EMR on EC2 that makes use of YARN as its useful resource supervisor. It hasn’t been examined on workloads from different frameworks that run on YARN, reminiscent of HIVE or TEZ.

Answer overview

The answer works by operating a Python script on the EMR cluster’s main node to gather metrics from the YARN useful resource supervisor and correlate them with price utilization particulars from the AWS Value and Utilization Experiences (AWS CUR). The script activated by a cronjob makes HTTP requests to the YARN useful resource supervisor to gather two forms of metrics from paths /ws/v1/cluster/metrics for cluster metrics and /ws/v1/cluster/apps for utility metrics. The cluster metrics comprise utilization info of cluster assets, and the appliance metrics comprise utilization info of an utility or job. These metrics are saved in an Amazon Easy Storage Service (Amazon S3) bucket.

There are two YARN metrics that seize the useful resource utilization info of an utility or job.

  • memorySeconds – That is the reminiscence (in MB) allotted to an utility occasions the variety of seconds the appliance ran
  • vcoreSeconds – That is the variety of YARN vcores allotted to an utility occasions the variety of seconds utility ran

The answer makes use of memorySeconds to derive the price of operating the appliance or job. It may be modified to make use of vcoreSeconds as an alternative if vital.

The metadata of the YARN metrics collected in Amazon S3 is created, saved, and represented as database and tables in AWS Glue Knowledge Catalog, which is in flip out there to Amazon Athena for additional processing. Now you can write SQL queries in Athena to correlate the YARN metrics with the fee utilization info from AWS CUR to derive the detailed price breakdown of your EMR cluster by infrastructure and utility. This resolution creates two corresponding Athena views of the respective price breakdown that can grow to be the info supply to Amazon QuickSight for visualization.

The next diagram exhibits the answer structure.

EMR Cluster Usage Utility Solution Architecture

Stipulations

To carry out the answer, you want the next conditions:

  1. Verify {that a} CUR is created in your AWS account. It wants an S3 bucket to retailer the report recordsdata. Observe the steps described in Creating Value and Utilization Experiences to create the CUR on the AWS Administration Console. When creating the report, make sure that the next settings are enabled:
    • Embody useful resource IDs
    • Time granularity is ready to hourly
    • Report information integration to Athena

It might probably take as much as 24 hours for AWS to start out delivering studies to your S3 bucket. Thereafter, your CUR will get up to date no less than one time a day.

  1. The answer wants Athena to run queries towards the info from the CUR utilizing normal SQL. To automate and streamline the combination of Athena with CUR, AWS supplies an AWS CloudFormation template, crawler-cfn.yml, which is mechanically generated in the identical S3 bucket throughout CUR creation. Observe the directions in Establishing Athena utilizing AWS CloudFormation templates to combine Athena with the CUR. This template will create an AWS Glue database that references to the CUR, an AWS Lambda occasion and an AWS Glue crawler that will get invoked by S3 occasion notification to replace the AWS Glue database each time the CUR will get up to date.
  2. Be sure to activate the AWS generated price allocation tag, aws:elasticmapreduce:job-flow-id. This allows the sphere, resource_tags_aws_elasticmapreduce_job_flow_id, within the CUR to be populated with the EMR cluster ID and is utilized by the SQL queries within the resolution. To activate the fee allocation tag from the administration console, observe these steps:
    • Check in to the payer account’s AWS Administration Console and open the AWS Billing and Value Administration console
    • Within the navigation pane, select Value Allocation Tags
    • Underneath AWS generated price allocation tags, select the aws:elasticmapreduce:job-flow-id tag
    • Select Activate. It might probably take as much as 24 hours for tags to activate.

The next screenshot exhibits an instance of the aws:elasticmapreduce:job-flow-id tag being activated.

CostAllocationTag

Now you can take a look at out this resolution on an EMR cluster in a lab surroundings. Should you’re not already conversant in EMR, observe the detailed directions offered in Tutorial: Getting began with Amazon EMR to launch a brand new EMR cluster and run a pattern Spark job.

Deploying the answer

To deploy the answer, observe the steps within the subsequent sections.

Putting in scripts to the EMR cluster

Obtain two scripts from the GitHub repository and save them into an S3 bucket:

  • emr_usage_report.py – Python script that makes the HTTP requests to YARN Useful resource Supervisor
  • emr_install_report.sh  – Bash script that creates a cronjob to run the python script each minute

To put in the scripts, add a step to the EMR cluster by the console or AWS Command Line Interface (AWS CLI) utilizing aws emr add-step command.

Change:

  • REGION with the AWS Areas the place the cluster is operating (for instance, Europe (Eire) eu-west-1)
  • MY-BUCKET with the title of the bucket the place the script is saved (for instance, my.artifact.bucket)
  • MY_REPORT_BUCKET with the bucket title the place you need to gather YARN metrics (for instance, my.report.bucket)
aws emr add-steps 
--cluster-id j-XXXXXXXXXXXXX 
--steps Kind=CUSTOM_JAR,Identify="Set up YARN reporter",Jar=s3://REGION.elasticmapreduce/libs/script-runner/script-runner.jar,Args=[s3://<MY-BUCKET>/emr-install_reporter.sh,s3://<MY-BUCKET>/emr_usage_reporter.py,MY_REPORT_BUCKET]

Now you can run some Spark jobs in your EMR cluster to start out producing utility utilization metrics.

Launching the CloudFormation stack

When the conditions are met and you’ve got the scripts deployed in order that your EMR clusters are sending YARN metrics to an S3 bucket, the remainder of the answer could be deployed utilizing CloudFormation.

Earlier than launching the stack, add a duplicate of this QuickSight definition file into an S3 bucket required by the CloudFormation template to construct the preliminary evaluation in QuickSight. When prepared, proceed to launch your stack to provision the remaining assets of the answer.

  1. Select

This mechanically launches AWS CloudFormation in your AWS account with a template. It prompts you to register as wanted and be sure you create the stack in your supposed Area.

The CloudFormation stack requires a couple of parameters, as proven within the following screenshot.

CloudFormationStack

The next desk describes the parameters.

ParameterDescription
Stack titleA significant title for the stack; for instance, EMRUsageReport
S3 configuration
YARNS3BucketNameIdentify of S3 bucket the place YARN metrics are saved
Value Utilization Report configuration
CURDatabaseNameIdentify of Value Utilization Report database in AWS Glue
CURTableNameIdentify of Value Utilization Report desk in AWS Glue
AWS Glue Database configuration
EMRUsageDBNameIdentify of AWS Glue database to be created for the EMR Value Utilization Report
EMRInfraTableNameIdentify of AWS Glue desk to be created for infrastructure utilization metrics
EMRAppTableNameIdentify of AWS Glue desk to be created for utility utilization metrics
QuickSight configuration
QSUserNameIdentify of QuickSight consumer in default namespace to handle the EMR Utilization Report assets in QuickSight.
QSDefinitionsFileS3 URI of the definition JSON file for the EMR Utilization Report.
  1. Enter the parameter values from the previous desk.
  2. Select Subsequent.
  3. On the subsequent display screen, enter any vital tags, an AWS Id and Entry Administration (IAM) position, stack failure, or superior choices if vital. In any other case, you possibly can go away them as default.
  4. Select Subsequent.
  5. Evaluate the small print on the ultimate display screen and choose the test containers confirming AWS CloudFormation may create IAM assets with customized names or require CAPABILITY_AUTO_EXPAND.
    CloudFormationCheckbox
  6. Select Create.

The stack will take a few minutes to create the remaining assets for the answer. After the CloudFormation stack is created, on the Outputs tab, you could find the small print of the assets created.

Reviewing the correlation outcomes

The CloudFormation template creates two Athena views containing the correlated price breakdown particulars of the YARN cluster and utility metrics with the CUR. The CUR aggregates price hourly and due to this fact correlation to derive the price of operating an utility is prorated primarily based on the hourly operating price of the EMR cluster.

The next screenshot exhibits the Athena view for the correlated price breakdown particulars of YARN cluster metrics.

CorrelationResults

The next desk describes the fields within the Athena view for YARN cluster metrics.

SubjectKindDescription
cluster_idstringID of the cluster.
householdstringUseful resource kind of the cluster. Attainable values are compute occasion, elastic map cut back occasion, storage and information switch.
billing_starttimestampBegin billing hour of the useful resource.
usage_typestringA selected kind or unit of the useful resource reminiscent of BoxUsage:m5.xlarge of compute occasion.
pricestringValue related to the useful resource.

The next screenshot exhibits the Athena view for the correlated price breakdown particulars of YARN utility metrics.

CostBreakdownYARNAppMetrics

The next desk describes the fields within the Athena view for YARN utility metrics.

SubjectKindDescription
cluster_idstringID of the cluster
idstringDistinctive identifier of the appliance run
consumerstringPerson title
titlestringIdentify of the appliance
queuestringQueue title from YARN useful resource supervisor
finalstatusstringLast standing of utility
applicationtypestringKind of the appliance
startedtimetimestampBegin time of the appliance
finishedtimetimestampFinish time of the appliance
elapsed_secdoubleTime taken to run the appliance
memorysecondsbigintThe reminiscence (in MB) allotted to an utility occasions the variety of seconds the appliance ran
vcoresecondsintThe variety of YARN vcores allotted to an utility occasions the variety of seconds utility ran
total_memory_mb_avgdoubleComplete quantity of reminiscence (in MB) out there to the cluster within the hour
memory_sec_costdoubleDerived unit price of memoryseconds
application_costdoubleDerived price related to the appliance primarily based on memoryseconds
total_costdoubleComplete price of assets related to the cluster for the hour

Constructing your personal visualization

In QuickSight, the CloudFormation template creates two datasets that reference Athena views as information sources and a pattern evaluation. The pattern evaluation has two sheets, EMR Infra Spend and EMR App Spend. They’ve a prepopulated bar chart and pivot tables to reveal how you need to use the datasets to construct your personal visualization to current the fee breakdown particulars of your EMR clusters.

EMR Infra Spend sheet references to the YARN cluster metrics dataset. There’s a filter for date vary choice and a filter for cluster ID choice. The pattern bar chart exhibits the consolidated price breakdown of the assets for every cluster in the course of the interval. The pivot desk breaks them down additional to point out their each day expenditure.

The next screenshot exhibits the EMR Infra Spend sheet from pattern evaluation created by the CloudFormation template.

EMR App Spend sheet references to the YARN utility metrics. There’s a filter for date vary choice and a filter for cluster ID choice. The pivot desk on this sheet exhibits how you need to use the fields within the dataset to current the fee breakdown particulars of the cluster by customers to watch the functions that had been run, whether or not they had been accomplished efficiently or not, the time and length of every run, and the derived price of the run.

The next screenshot exhibits the EMR App Spend sheet from pattern evaluation created by the CloudFormation template.

Cleanup

Should you now not want the assets you created throughout this walkthrough, delete them to forestall incurring further prices. To wash up your assets, full the next steps:

  1. On the CloudFormation console, delete the stack that you just created utilizing the template
  2. Terminate the EMR cluster
  3. Empty or delete the S3 bucket used for YARN metrics

Conclusion

On this submit, we mentioned tips on how to implement a complete cluster utilization reporting resolution that gives granular visibility into the useful resource consumption and related prices of particular person functions operating in your Amazon EMR on EC2 cluster. By utilizing the facility of Athena and QuickSight to correlate YARN metrics with price utilization particulars out of your Value and Utilization Report, this resolution empowers organizations to make knowledgeable selections. With these insights, you possibly can optimize useful resource allocation, implement honest and clear billing fashions primarily based on precise utility utilization, and finally obtain larger cost-efficiency in your EMR environments. This resolution will enable you unlock the total potential of your EMR cluster, driving steady enchancment in your information processing and analytics workflows whereas maximizing return on funding.


In regards to the authors

Boon Lee Eu is a Senior Technical Account Supervisor at Amazon Net Companies (AWS). He works intently and proactively with Enterprise Help clients to offer advocacy and strategic technical steerage to assist plan and obtain operational excellence in AWS surroundings primarily based on greatest practices. Based mostly in Singapore, Boon Lee has over 20 years of expertise in IT & Telecom industries.

Kyara Labrador is a Sr. Analytics Specialist Options Architect at Amazon Net Companies (AWS) Philippines, specializing in large information and analytics. She helps clients in designing and implementing scalable, safe, and cost-effective information options, in addition to migrating and modernizing their large information and analytics workloads to AWS. She is obsessed with empowering organizations to unlock the total potential of their information.

Vikas Omer is the Head of Knowledge & AI Answer Structure for ASEAN at Amazon Net Companies (AWS). With over 15 years of expertise within the information and AI house, he’s a seasoned chief who leverages his experience to drive innovation and growth within the area. Vikas is obsessed with serving to clients and companions succeed of their digital transformation journeys, specializing in cloud-based options and rising applied sciences.

Lorenzo Ripani is a Massive Knowledge Answer Architect at AWS. He’s obsessed with distributed methods, open supply applied sciences and safety. He spends most of his time working with clients all over the world to design, consider and optimize scalable and safe information pipelines with Amazon EMR.

Related Articles

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Latest Articles