© 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved
Building an Analytics Pipeline with a small team.
Mark Bate,
Solutions Architect (Games)
Redshift
Data warehousing
EMR
Hadoop + Spark
Athena
Interactive analytics
Kinesis Data Analytics
Real time
Elasticsearch Service
Operational Analytics
RDS
MySQL, PostgreSQL, MariaDB,
Oracle, SQL Server
Aurora
MySQL, PostgreSQL
API-driven AI ServicesQuickSight SageMaker
DynamoDB
Key value
ElastiCache
Redis, Memcached
Neptune
Graph
Timestream
Time Series
QLDB
Ledger Database
S3/ Glacier
Glue
ETL & Data Catalog
Lake Formation
Data Lakes
Database Migration Service | Snowball | Snowmobile | Kinesis Data Firehose | Kinesis Data Streams | Managed Streaming for Kafka
Data Movement
Analytics Databases
Business Intelligence & Machine Learning
Data Lake
Managed
Blockchain
Blockchain
Templates
Blockchain
RDS on VMware
DocumentDB
Document
AWS analytics portfolio
Broad and deep portfolio, purpose-built for builders
Attract more players
Increase their enjoyment
Longer play sessions
Encourage engagement
“Analytics can’t make a hit game -
but can improve it!”
Analytics flywheel for games
Timely decisions require new data in
minutes
Real time Seconds Minutes Hours Days Months
Valueofdatatodecision-making
Preventive/Predictive
Actionable Reactive Historical
Time critical decisions Traditional “batch” business intelligence
Data loses value quickly over time
Source: Perishable insights, Mike Gualtieri, Forrester
How do I build content players truly enjoy?
Know what players
are doing and when
• Sessions
• Progress
• Location
• Blockers
• Feedback
Measure and
validate core game
mechanics
• Gameplay
• Player choices
• Goals
• Performance
Experiment
• Scientific method
• Validate feedback
Analytic: Level design
Analytic: Level design
Telemetry
(events)
Metrics
(answers)
Anatomy of an analytics pipeline
Telemetry
(events)
{
‘event_id’ : ‘05b00439-6a07-4112-9c8d-165f1643e5d1’,
‘event_type’ : ‘player_death’,
‘event_timestamp’ : ‘2017-05-01T21:05:18.000Z’,
‘arena’ : ‘el_dorado’,
‘position_x’ : 507.12,
‘position_y’ : 551.61
}
Ingest/
Collect
Store
Process/
analyze
Consume/
visualize
Metrics
(answers)
Avoid wasted effort here
Not adding differentiated value
Traditionally difficult to scale
Unpredictable utilization
Anatomy of analytics pipelines
Analytics pipeline behind Fortnite
Game
clients
Game
servers
Launcher
Game
services
Grafana
Scoreboards API
Limited raw data
(real time ad-hoc SQL)User ETL
(metric definition)
Spark on EMR Amazon
DynamoDB
NEAR REAL-TIME
PIPELINE
BATCH PIPELINE
ETL
using EMR
Tableau/BI
Ad-hoc SQLS3
(Data lake)
Amazon
Kinesis
APIs
Databases
S3
Other
sources
Analytics platform runs onAWS
125M events/min ingested at peak
Amazon S3 leveraged as a data lake, >35 petabytes
Amazon Kinesis used to collect telemetry data
Real-time analytics done through Spark on Amazon EMR
Amazon EMR for large batch data processing
Game designers use data to inform their decisions
Streaming with Amazon Kinesis
SQL
Analyze data streams with SQL or
Java (Flink)
Kinesis Data Analytics
Capture, process, and store data
streams
Kinesis Data Streams
Load data streams into data
stores
Kinesis Data Firehose
Easily collect, process, and analyze data and video streams in real time
Streaming real-time data with AWS
Hadoop with Amazon EMR
Your own code on Amazon EC2
Centralized Data lake / Single
source of truth /
Data Warehouse
(Amazon S3,
Amazon Redshift)
Ingest
data streams
Kinesis Data
Streams
Kinesis Data
Analytics
Analyze data
streams with SQL
or Java
Kinesis Data
Firehose
Load into
data stores
Serverless w/ AWS Lambda
Amazon Kinesis for analytics pipelines
Collect, process, and analyze data streams in real time
Flexible
processing options
Batch Layer
Game
Backend
Amazon
Kinesis Data
Firehose
Game
Clients
Visualization
Dashboard
Amazon S3 Amazon AthenaAWS Glue
Speed Layer
Amazon
Kinesis Data
Firehose
Game
Clients
Amazon
Elasticsearch
Service
Kibana
Dashboard
Game
Backend
Authentication options
Reference: Serverless Real-Time Analytics for Mobile Gaming
Get started!
Gaming Analytics Pipeline Solution
• Reference architecture with pre-built templates,
instructions
• Launch into your account with AWS CloudFormation
• Learn more:
https://aws.amazon.com/solutions/gaming-analytics-
pipeline/
• Reference:
https://d1.awsstatic.com/architecture-
diagrams/ArchitectureDiagrams/serverless-analytics-
for-mobile-gaming.pdf

Game Analytics Nov 2019

  • 1.
    © 2019, AmazonWeb Services, Inc. or its Affiliates. All rights reserved Building an Analytics Pipeline with a small team. Mark Bate, Solutions Architect (Games)
  • 2.
    Redshift Data warehousing EMR Hadoop +Spark Athena Interactive analytics Kinesis Data Analytics Real time Elasticsearch Service Operational Analytics RDS MySQL, PostgreSQL, MariaDB, Oracle, SQL Server Aurora MySQL, PostgreSQL API-driven AI ServicesQuickSight SageMaker DynamoDB Key value ElastiCache Redis, Memcached Neptune Graph Timestream Time Series QLDB Ledger Database S3/ Glacier Glue ETL & Data Catalog Lake Formation Data Lakes Database Migration Service | Snowball | Snowmobile | Kinesis Data Firehose | Kinesis Data Streams | Managed Streaming for Kafka Data Movement Analytics Databases Business Intelligence & Machine Learning Data Lake Managed Blockchain Blockchain Templates Blockchain RDS on VMware DocumentDB Document AWS analytics portfolio Broad and deep portfolio, purpose-built for builders
  • 3.
    Attract more players Increasetheir enjoyment Longer play sessions Encourage engagement “Analytics can’t make a hit game - but can improve it!” Analytics flywheel for games
  • 4.
    Timely decisions requirenew data in minutes Real time Seconds Minutes Hours Days Months Valueofdatatodecision-making Preventive/Predictive Actionable Reactive Historical Time critical decisions Traditional “batch” business intelligence Data loses value quickly over time Source: Perishable insights, Mike Gualtieri, Forrester
  • 5.
    How do Ibuild content players truly enjoy? Know what players are doing and when • Sessions • Progress • Location • Blockers • Feedback Measure and validate core game mechanics • Gameplay • Player choices • Goals • Performance Experiment • Scientific method • Validate feedback
  • 6.
  • 7.
  • 8.
  • 9.
    Telemetry (events) { ‘event_id’ : ‘05b00439-6a07-4112-9c8d-165f1643e5d1’, ‘event_type’: ‘player_death’, ‘event_timestamp’ : ‘2017-05-01T21:05:18.000Z’, ‘arena’ : ‘el_dorado’, ‘position_x’ : 507.12, ‘position_y’ : 551.61 } Ingest/ Collect Store Process/ analyze Consume/ visualize Metrics (answers) Avoid wasted effort here Not adding differentiated value Traditionally difficult to scale Unpredictable utilization Anatomy of analytics pipelines
  • 10.
    Analytics pipeline behindFortnite Game clients Game servers Launcher Game services Grafana Scoreboards API Limited raw data (real time ad-hoc SQL)User ETL (metric definition) Spark on EMR Amazon DynamoDB NEAR REAL-TIME PIPELINE BATCH PIPELINE ETL using EMR Tableau/BI Ad-hoc SQLS3 (Data lake) Amazon Kinesis APIs Databases S3 Other sources Analytics platform runs onAWS 125M events/min ingested at peak Amazon S3 leveraged as a data lake, >35 petabytes Amazon Kinesis used to collect telemetry data Real-time analytics done through Spark on Amazon EMR Amazon EMR for large batch data processing Game designers use data to inform their decisions
  • 11.
    Streaming with AmazonKinesis SQL Analyze data streams with SQL or Java (Flink) Kinesis Data Analytics Capture, process, and store data streams Kinesis Data Streams Load data streams into data stores Kinesis Data Firehose Easily collect, process, and analyze data and video streams in real time
  • 12.
  • 13.
    Hadoop with AmazonEMR Your own code on Amazon EC2 Centralized Data lake / Single source of truth / Data Warehouse (Amazon S3, Amazon Redshift) Ingest data streams Kinesis Data Streams Kinesis Data Analytics Analyze data streams with SQL or Java Kinesis Data Firehose Load into data stores Serverless w/ AWS Lambda Amazon Kinesis for analytics pipelines Collect, process, and analyze data streams in real time Flexible processing options
  • 14.
  • 15.
  • 16.
    Authentication options Reference: ServerlessReal-Time Analytics for Mobile Gaming
  • 17.
    Get started! Gaming AnalyticsPipeline Solution • Reference architecture with pre-built templates, instructions • Launch into your account with AWS CloudFormation • Learn more: https://aws.amazon.com/solutions/gaming-analytics- pipeline/ • Reference: https://d1.awsstatic.com/architecture- diagrams/ArchitectureDiagrams/serverless-analytics- for-mobile-gaming.pdf

Editor's Notes

  • #14 Data lake is a whole other discussion but basically comes down to Cold/warm/hot data