SlideShare une entreprise Scribd logo
1  sur  27
How to Guarantee Exact COUNT DISTINCT Queries
with Sub-Second Latency on Massive Datasets
Kaige Liu
2020.5
© Kyligence Inc. 2019, Confidential.
Business Scenarios
Technical Principles
Demo
Use Cases
Q&A
Agenda
© Kyligence Inc. 2019, Confidential.
Business Scenarios
© Kyligence Inc. 2019, Confidential.
What Is Count Distinct?
Count Distinct is used to compute the number of
unique values in a data set.
• PV (Page View)
• UV (Unique Visitors)
ID Username Page
1 Alice /kyligence
2 Alice /Kyligence/Blog
3 Carol /Kyligence/Events
4 Bob /Kyligence/Resources
5 Alice /Kyligence/Downloads
Alice, Bob, Carol
3
© Kyligence Inc. 2019, Confidential.
Approximate and Exact Count Distinct
• Approximate Count Distinct
• Quick, less memory/CPU
• Not accurate
• Trend analysis, small errors are acceptable
• Exact Count Distinct
• Slow, more memory/CPU
• Accurate
• Transaction relevant. Paid Advertising, Precision Marketing, etc.
Error Rate $ 1 Million $ 1 Billion
1.22% $12,200 $12,200,000
2.44% $24,000 $24,000,000
9.75% $97,500 $97,500,000
© Kyligence Inc. 2019, Confidential.
Where
are they
coming
from?
Who are
my
visitors?
Web/Ap
p
Analytic
s
Which
page lost
the most
users?
How
many
active
users?
How
many
new
users?
How
many
unique
visitors?
Scenarios - Web/App Analytics
© Kyligence Inc. 2019, Confidential.
Scenarios - User Behavior Analytics
Retention Analysis
Funnel Analysis
© Kyligence Inc. 2019, Confidential.
Technical Principles
© Kyligence Inc. 2019, Confidential.
Challenges with Exact Count Distinct
• Approximate Count Distinct is easy – HyperLogLog
• Exact Count Distinct is a big challenge for all query engines at massive scale
Challenges
• Bad performance – Need to scan all data
• Non-cumulative – Hard to do rollup and/or operations
• Hard to optimize on multiple columns
• Analysis always requires more than one count distinct operation
© Kyligence Inc. 2019, Confidential.
Count Distinct Performance on Different Platforms
• Google BigQuery
• Snowflake
• Athena
• Apache Kylin
• Kyligence
© Kyligence Inc. 2019, Confidential.
Kyligence = Kylin + Intelligence
• Founded in 2016 by the creators of Apache Kylin
• Built around Kylin, with augmented AI and enhanced to deliver
unprecedented enterprise analytic performance
• CRN Top-10 big data startups in 2018
• Global Presence: San Jose, Seattle, New York, Shanghai, Beijing
• VCs: Fidelity International, Shunwei Capital, Broadband Capital,
Redpoint, Cisco, Coatue
Accelerate Critical Business Decisions with AI-Augmented Data Management
and Analytics
2016
Founded Pre-
A
Redpoint
Cisco
2017
Series A
CBC
Shunwei
2018
Series B
8Roads
2019
Series C
Coatue
© Kyligence Inc. 2019, Confidential.
How Does Apache Kylin Achieve This?
BitmapPre-Aggregation
• Pre-aggregate count distinct in cubes
• Fetch results directly without on the
fly calculations
• Supports Rollup
• Reduces memory/storage significantly
• Supports String type and detail queries
Dictionary
© Kyligence Inc. 2019, Confidential.
Pre-Aggregation
Date UID Page
2020-04-01
01
1 /kyligence
2020-04-01
01
1 /Kyligence/Blog
2020-04-01
01
2 /Kyligence/News
2020-04-02
02
3 /Kyligence/Events
2020-04-02
02
2 /Kyligence/Resources
2020-04-02
02
1 /Kyligence/Downloads
Date Count(UID) Count(distinct UID)
UID)
2020-04-01
01
3 2
2020-04-02
02
3 3
Date Count(UID) Count(distinct UID)
2020-04-01
01 and
2020-04-02
02
6 ??
© Kyligence Inc. 2019, Confidential.
7 6 5 4 3 2 1 0
Bitmap
UID
1
2
4
5
7
9
10
11
13
1 0 0 1 0 1 1 0
0 0 1 0 1 1 1 0
Table Bitmap
• Saves storage significantly
• Supports logical operations directly
• Contains information needed to do
aggregation
• RoaringBitmap
© Kyligence Inc. 2019, Confidential.
Bitmap
Date UID Page
2020-04-01
01
1 /kyligence
2020-04-01
01
1 /Kyligence/Blog
2020-04-01
01
2 /Kyligence/News
2020-04-02
02
3 /Kyligence/Events
2020-04-02
02
2 /Kyligence/Resources
2020-04-02
02
1 /Kyligence/Downloads
Date Count(UID) Count(distinct UID)
UID)
2020-04-01
01
3 2
2020-04-02
02
3 3
Date Count(UID) Count(distinct UID)
2020-04-01
01 and
2020-04-02
02
6 Bitmap(1,2,3)
Date Count(UID) Count(distinct UID)
UID)
2020-04-01 3 Bitmap(1,2)
2020-04-02 3 Bitmap(1,2,3)
© Kyligence Inc. 2019, Confidential.
Operations in Bitmap
• Two bitmaps, each containing two different data sets:
[1, 3, 4, 5]
[2, 3, 4, 6]
• And - All elements contained in both bitmaps:
[1, 3, 4, 5] and [2, 3, 4, 6] = [3, 4]
Scenarios: Retention Analysis, Funnel Analysis
• Or – All elements in either bitmap:
[1, 3, 4, 5] or [2, 3, 4, 6] = [1, 2, 3, 4, 5, 6]
Scenarios: Cross-Dimension Analysis
© Kyligence Inc. 2019, Confidential.
Dictionary
Date USERNAME Page
2020-04-01
01
Alice /kyligence
2020-04-01
01
Alice /Kyligence/Blog
2020-04-01
01
Bob /Kyligence/News
2020-04-02
02
Coral /Kyligence/Events
2020-04-02
02
Bob /Kyligence/Resources
2020-04-02
02
Alice /Kyligence/Downloads
USERNAME ECODED
Alice 1
Bob 2
Coral 3
Date Count(UID) Count(distinct UID)
2020-04-01
01 and
2020-04-02
02
6 Bitmap(1,2,3)
Date Count(UID) Count(distinct UID)
UID)
2020-04-01
01
3 Bitmap(1,2)
2020-04-02
02
3 Bitmap(1,2,3)
Bitmap can only support int values. How about String columns?
Dictionary
© Kyligence Inc. 2019, Confidential.
Use Cases
© Kyligence Inc. 2019, Confidential.
Manbang Group
• The largest Chinese truck logistics startup
• 7 million+ trucks
• 2.25 million active users
• 8 apps and 10 TB+ data
Requirements
• Retention analysis on a wide range of dimensions
and date ranges
• Funnel analysis with ability to customize funnel
• User profile analysis
© Kyligence Inc. 2019, Confidential.
Architecture with Apache Kylin
© Kyligence Inc. 2019, Confidential.
Retention Analysis for Manbang Group
• Users can choose any column and any date range to do the retention analysis
© Kyligence Inc. 2019, Confidential.
Funnel Analysis for Manbang group
• Users can customize funnels with any number of steps
• Can identify the specific users lost between steps
© Kyligence Inc. 2019, Confidential.
DiDi
• #1 ride-share company in China
• 92 million monthly active users
(as of Dec. 2019)
• 24 million rides per day in 2019
Requirements
• User profile analysis
• Precision marketing
© Kyligence Inc. 2019, Confidential.
Scenarios – Apache Kylin in Didi
• Precision Marketing
o Send coupons to exact target users
o Upgrade cars for specific users
• Promotion Activity Analysis
o How many new/returned users are gained in this activity?
o Which kind of users are most interested in this activity?
• Optimize User Experience
o Which stages lost the most users?
o How to increase customer stickiness?
User Profile
Precision
Marketing
User
Behavior
Analysis
User Tags
Workflow
Analysis
Promotion
Activity
Analysis
© Kyligence Inc. 2019, Confidential.
Didi Kylin Usage
200 TB+ 5,000+ 7,000+ 7
Data Cubes Jobs per day Clusters
© Kyligence Inc. 2019, Confidential.
Join the Community
https://github.com/apache/kylin apache-kylin.slack.comuser@kylin.apache.org
THANK YOU

Contenu connexe

Tendances

Logicalis IoT & Smart Cities (Use Case)
Logicalis IoT & Smart Cities (Use Case)Logicalis IoT & Smart Cities (Use Case)
Logicalis IoT & Smart Cities (Use Case)Cloudera, Inc.
 
Ayush Tiwari [PTC] | Unlock IoT Value with PTC’s ThingWorx Platform & InfluxD...
Ayush Tiwari [PTC] | Unlock IoT Value with PTC’s ThingWorx Platform & InfluxD...Ayush Tiwari [PTC] | Unlock IoT Value with PTC’s ThingWorx Platform & InfluxD...
Ayush Tiwari [PTC] | Unlock IoT Value with PTC’s ThingWorx Platform & InfluxD...InfluxData
 
How to Gain a Competitive Edge with an Open Source, Purpose-built Time Series...
How to Gain a Competitive Edge with an Open Source, Purpose-built Time Series...How to Gain a Competitive Edge with an Open Source, Purpose-built Time Series...
How to Gain a Competitive Edge with an Open Source, Purpose-built Time Series...DevOps.com
 
Will Edge Computing IoT Solutions be a Real Trend in 2019?
Will Edge Computing IoT Solutions be a Real Trend in 2019?Will Edge Computing IoT Solutions be a Real Trend in 2019?
Will Edge Computing IoT Solutions be a Real Trend in 2019?Tyrone Systems
 
Enabling Push Button Productization of AI Models
Enabling Push Button Productization of AI ModelsEnabling Push Button Productization of AI Models
Enabling Push Button Productization of AI ModelsDatabricks
 
Data Science in the Enterprise
Data Science in the EnterpriseData Science in the Enterprise
Data Science in the EnterpriseThe Hive
 
The Hive Think Tank: Talk by Mohandas Pai - India at 2030, How Tech Entrepren...
The Hive Think Tank: Talk by Mohandas Pai - India at 2030, How Tech Entrepren...The Hive Think Tank: Talk by Mohandas Pai - India at 2030, How Tech Entrepren...
The Hive Think Tank: Talk by Mohandas Pai - India at 2030, How Tech Entrepren...The Hive
 
Cloud Computing
Cloud ComputingCloud Computing
Cloud Computingayushi19
 
FIWARE Global Summit - Edge/Fog Computing in “Powered by FIWARE” Architectures
FIWARE Global Summit - Edge/Fog Computing in “Powered by FIWARE” ArchitecturesFIWARE Global Summit - Edge/Fog Computing in “Powered by FIWARE” Architectures
FIWARE Global Summit - Edge/Fog Computing in “Powered by FIWARE” ArchitecturesFIWARE
 
Data Natives Frankfurt v 11.0 | "Competitive advantages with knowledge graphs...
Data Natives Frankfurt v 11.0 | "Competitive advantages with knowledge graphs...Data Natives Frankfurt v 11.0 | "Competitive advantages with knowledge graphs...
Data Natives Frankfurt v 11.0 | "Competitive advantages with knowledge graphs...Dataconomy Media
 
OpenPOWER partner presentation - GTS Data
OpenPOWER partner presentation - GTS DataOpenPOWER partner presentation - GTS Data
OpenPOWER partner presentation - GTS DataGanesan Narayanasamy
 
Visualizing Big Data with augmented and virtual reality
Visualizing Big Data with augmented and virtual realityVisualizing Big Data with augmented and virtual reality
Visualizing Big Data with augmented and virtual realityMolham Al-Maleh
 
InfoSphere Optim archive for archive/purge of application data
InfoSphere Optim archive for archive/purge of application dataInfoSphere Optim archive for archive/purge of application data
InfoSphere Optim archive for archive/purge of application dataBharath Nunepalli
 
This Week in Data Science - Top 5 News - April 26, 2019
This Week in Data Science - Top 5 News - April 26, 2019This Week in Data Science - Top 5 News - April 26, 2019
This Week in Data Science - Top 5 News - April 26, 2019NVIDIA
 
Seven Ways to Boost Artificial Intelligence Research
Seven Ways to Boost Artificial Intelligence ResearchSeven Ways to Boost Artificial Intelligence Research
Seven Ways to Boost Artificial Intelligence ResearchNVIDIA
 
FIWARE Global Summit - Standard Data Models for the Integration of FIWARE and...
FIWARE Global Summit - Standard Data Models for the Integration of FIWARE and...FIWARE Global Summit - Standard Data Models for the Integration of FIWARE and...
FIWARE Global Summit - Standard Data Models for the Integration of FIWARE and...FIWARE
 
Has serverless adoption hit a roadblock?
Has serverless adoption hit a roadblock?Has serverless adoption hit a roadblock?
Has serverless adoption hit a roadblock?Veselin Pizurica
 
Create your Big Data vision and Hadoop-ify your data warehouse
Create your Big Data vision and Hadoop-ify your data warehouseCreate your Big Data vision and Hadoop-ify your data warehouse
Create your Big Data vision and Hadoop-ify your data warehouseJeff Kelly
 

Tendances (20)

Logicalis IoT & Smart Cities (Use Case)
Logicalis IoT & Smart Cities (Use Case)Logicalis IoT & Smart Cities (Use Case)
Logicalis IoT & Smart Cities (Use Case)
 
Ayush Tiwari [PTC] | Unlock IoT Value with PTC’s ThingWorx Platform & InfluxD...
Ayush Tiwari [PTC] | Unlock IoT Value with PTC’s ThingWorx Platform & InfluxD...Ayush Tiwari [PTC] | Unlock IoT Value with PTC’s ThingWorx Platform & InfluxD...
Ayush Tiwari [PTC] | Unlock IoT Value with PTC’s ThingWorx Platform & InfluxD...
 
How to Gain a Competitive Edge with an Open Source, Purpose-built Time Series...
How to Gain a Competitive Edge with an Open Source, Purpose-built Time Series...How to Gain a Competitive Edge with an Open Source, Purpose-built Time Series...
How to Gain a Competitive Edge with an Open Source, Purpose-built Time Series...
 
Will Edge Computing IoT Solutions be a Real Trend in 2019?
Will Edge Computing IoT Solutions be a Real Trend in 2019?Will Edge Computing IoT Solutions be a Real Trend in 2019?
Will Edge Computing IoT Solutions be a Real Trend in 2019?
 
Enabling Push Button Productization of AI Models
Enabling Push Button Productization of AI ModelsEnabling Push Button Productization of AI Models
Enabling Push Button Productization of AI Models
 
Data Science in the Enterprise
Data Science in the EnterpriseData Science in the Enterprise
Data Science in the Enterprise
 
The Hive Think Tank: Talk by Mohandas Pai - India at 2030, How Tech Entrepren...
The Hive Think Tank: Talk by Mohandas Pai - India at 2030, How Tech Entrepren...The Hive Think Tank: Talk by Mohandas Pai - India at 2030, How Tech Entrepren...
The Hive Think Tank: Talk by Mohandas Pai - India at 2030, How Tech Entrepren...
 
Cloud Computing
Cloud ComputingCloud Computing
Cloud Computing
 
FIWARE Global Summit - Edge/Fog Computing in “Powered by FIWARE” Architectures
FIWARE Global Summit - Edge/Fog Computing in “Powered by FIWARE” ArchitecturesFIWARE Global Summit - Edge/Fog Computing in “Powered by FIWARE” Architectures
FIWARE Global Summit - Edge/Fog Computing in “Powered by FIWARE” Architectures
 
Data Natives Frankfurt v 11.0 | "Competitive advantages with knowledge graphs...
Data Natives Frankfurt v 11.0 | "Competitive advantages with knowledge graphs...Data Natives Frankfurt v 11.0 | "Competitive advantages with knowledge graphs...
Data Natives Frankfurt v 11.0 | "Competitive advantages with knowledge graphs...
 
OpenPOWER partner presentation - GTS Data
OpenPOWER partner presentation - GTS DataOpenPOWER partner presentation - GTS Data
OpenPOWER partner presentation - GTS Data
 
Visualizing Big Data with augmented and virtual reality
Visualizing Big Data with augmented and virtual realityVisualizing Big Data with augmented and virtual reality
Visualizing Big Data with augmented and virtual reality
 
InfoSphere Optim archive for archive/purge of application data
InfoSphere Optim archive for archive/purge of application dataInfoSphere Optim archive for archive/purge of application data
InfoSphere Optim archive for archive/purge of application data
 
This Week in Data Science - Top 5 News - April 26, 2019
This Week in Data Science - Top 5 News - April 26, 2019This Week in Data Science - Top 5 News - April 26, 2019
This Week in Data Science - Top 5 News - April 26, 2019
 
Seven Ways to Boost Artificial Intelligence Research
Seven Ways to Boost Artificial Intelligence ResearchSeven Ways to Boost Artificial Intelligence Research
Seven Ways to Boost Artificial Intelligence Research
 
AI at the Edge
AI at the EdgeAI at the Edge
AI at the Edge
 
FIWARE Global Summit - Standard Data Models for the Integration of FIWARE and...
FIWARE Global Summit - Standard Data Models for the Integration of FIWARE and...FIWARE Global Summit - Standard Data Models for the Integration of FIWARE and...
FIWARE Global Summit - Standard Data Models for the Integration of FIWARE and...
 
Has serverless adoption hit a roadblock?
Has serverless adoption hit a roadblock?Has serverless adoption hit a roadblock?
Has serverless adoption hit a roadblock?
 
Create your Big Data vision and Hadoop-ify your data warehouse
Create your Big Data vision and Hadoop-ify your data warehouseCreate your Big Data vision and Hadoop-ify your data warehouse
Create your Big Data vision and Hadoop-ify your data warehouse
 
Opportunities derived by AI
Opportunities derived by AIOpportunities derived by AI
Opportunities derived by AI
 

Similaire à How to Guarantee Exact COUNT DISTINCT Queries with Sub-Second Latency on Massive Datasets

Take the Bias out of Big Data Insights With Augmented Analytics
Take the Bias out of Big Data Insights With Augmented AnalyticsTake the Bias out of Big Data Insights With Augmented Analytics
Take the Bias out of Big Data Insights With Augmented AnalyticsTyler Wishnoff
 
Augmented OLAP Analytics for Big Data
Augmented OLAP Analytics for Big DataAugmented OLAP Analytics for Big Data
Augmented OLAP Analytics for Big DataTyler Wishnoff
 
Augmented OLAP for Big Data
Augmented OLAP for Big DataAugmented OLAP for Big Data
Augmented OLAP for Big DataLuke Han
 
Simplify Data Analytics Over the Cloud
Simplify Data Analytics Over the CloudSimplify Data Analytics Over the Cloud
Simplify Data Analytics Over the CloudTyler Wishnoff
 
Legacy IBM Systems and Splunk: Security, Compliance and Uptime
Legacy IBM Systems and Splunk: Security, Compliance and UptimeLegacy IBM Systems and Splunk: Security, Compliance and Uptime
Legacy IBM Systems and Splunk: Security, Compliance and UptimePrecisely
 
The Enabling Power of Distributed SQL for Enterprise Digital Transformation I...
The Enabling Power of Distributed SQL for Enterprise Digital Transformation I...The Enabling Power of Distributed SQL for Enterprise Digital Transformation I...
The Enabling Power of Distributed SQL for Enterprise Digital Transformation I...NuoDB
 
Snowflake: The Good, the Bad, and the Ugly
Snowflake: The Good, the Bad, and the UglySnowflake: The Good, the Bad, and the Ugly
Snowflake: The Good, the Bad, and the UglyTyler Wishnoff
 
Snowflake: The Good, the Bad and the Ugly
Snowflake: The Good, the Bad and the UglySnowflake: The Good, the Bad and the Ugly
Snowflake: The Good, the Bad and the UglySamanthaBerlant
 
Lightning-Fast, Interactive Business Intelligence Performance with MicroStrat...
Lightning-Fast, Interactive Business Intelligence Performance with MicroStrat...Lightning-Fast, Interactive Business Intelligence Performance with MicroStrat...
Lightning-Fast, Interactive Business Intelligence Performance with MicroStrat...Tyler Wishnoff
 
Augmented OLAP for Big Data Analytics
Augmented OLAP for Big Data AnalyticsAugmented OLAP for Big Data Analytics
Augmented OLAP for Big Data AnalyticsTyler Wishnoff
 
Addressing the systemic shortcomings of cloud analytics
Addressing the systemic shortcomings of cloud analyticsAddressing the systemic shortcomings of cloud analytics
Addressing the systemic shortcomings of cloud analyticsSamanthaBerlant
 
Apache Kylin and Use Cases - 2018 Big Data Spain
Apache Kylin and Use Cases - 2018 Big Data SpainApache Kylin and Use Cases - 2018 Big Data Spain
Apache Kylin and Use Cases - 2018 Big Data SpainLuke Han
 
Ian Uriarte Timbergrove at IBM IoTExchange 2019
Ian Uriarte Timbergrove at IBM IoTExchange 2019Ian Uriarte Timbergrove at IBM IoTExchange 2019
Ian Uriarte Timbergrove at IBM IoTExchange 2019IanUriarte2
 
Cisco Connect Toronto 2018 an introduction to Cisco kinetic
Cisco Connect Toronto 2018   an introduction to Cisco kineticCisco Connect Toronto 2018   an introduction to Cisco kinetic
Cisco Connect Toronto 2018 an introduction to Cisco kineticCisco Canada
 
Cisco Connect Toronto 2018 an introduction to Cisco kinetic
Cisco Connect Toronto 2018   an introduction to Cisco kineticCisco Connect Toronto 2018   an introduction to Cisco kinetic
Cisco Connect Toronto 2018 an introduction to Cisco kineticCisco Canada
 
IoT Update | Hoe implementeer je IoT Schaalbaar in je IT landschap
IoT Update | Hoe implementeer je IoT Schaalbaar in je IT landschapIoT Update | Hoe implementeer je IoT Schaalbaar in je IT landschap
IoT Update | Hoe implementeer je IoT Schaalbaar in je IT landschapIoT Academy
 
IBM CDS Overview
IBM CDS OverviewIBM CDS Overview
IBM CDS OverviewJean Tan
 
The value of a connected factory
The value of a connected factoryThe value of a connected factory
The value of a connected factoryCroonwolter&dros
 
A Connected Data Landscape: Virtualization and the Internet of Things
A Connected Data Landscape: Virtualization and the Internet of ThingsA Connected Data Landscape: Virtualization and the Internet of Things
A Connected Data Landscape: Virtualization and the Internet of ThingsInside Analysis
 
Customer Presentation - IBM Cloud Pak for Data Overview (Level 100).PPTX
Customer Presentation - IBM Cloud Pak for Data Overview (Level 100).PPTXCustomer Presentation - IBM Cloud Pak for Data Overview (Level 100).PPTX
Customer Presentation - IBM Cloud Pak for Data Overview (Level 100).PPTXtsigitnist02
 

Similaire à How to Guarantee Exact COUNT DISTINCT Queries with Sub-Second Latency on Massive Datasets (20)

Take the Bias out of Big Data Insights With Augmented Analytics
Take the Bias out of Big Data Insights With Augmented AnalyticsTake the Bias out of Big Data Insights With Augmented Analytics
Take the Bias out of Big Data Insights With Augmented Analytics
 
Augmented OLAP Analytics for Big Data
Augmented OLAP Analytics for Big DataAugmented OLAP Analytics for Big Data
Augmented OLAP Analytics for Big Data
 
Augmented OLAP for Big Data
Augmented OLAP for Big DataAugmented OLAP for Big Data
Augmented OLAP for Big Data
 
Simplify Data Analytics Over the Cloud
Simplify Data Analytics Over the CloudSimplify Data Analytics Over the Cloud
Simplify Data Analytics Over the Cloud
 
Legacy IBM Systems and Splunk: Security, Compliance and Uptime
Legacy IBM Systems and Splunk: Security, Compliance and UptimeLegacy IBM Systems and Splunk: Security, Compliance and Uptime
Legacy IBM Systems and Splunk: Security, Compliance and Uptime
 
The Enabling Power of Distributed SQL for Enterprise Digital Transformation I...
The Enabling Power of Distributed SQL for Enterprise Digital Transformation I...The Enabling Power of Distributed SQL for Enterprise Digital Transformation I...
The Enabling Power of Distributed SQL for Enterprise Digital Transformation I...
 
Snowflake: The Good, the Bad, and the Ugly
Snowflake: The Good, the Bad, and the UglySnowflake: The Good, the Bad, and the Ugly
Snowflake: The Good, the Bad, and the Ugly
 
Snowflake: The Good, the Bad and the Ugly
Snowflake: The Good, the Bad and the UglySnowflake: The Good, the Bad and the Ugly
Snowflake: The Good, the Bad and the Ugly
 
Lightning-Fast, Interactive Business Intelligence Performance with MicroStrat...
Lightning-Fast, Interactive Business Intelligence Performance with MicroStrat...Lightning-Fast, Interactive Business Intelligence Performance with MicroStrat...
Lightning-Fast, Interactive Business Intelligence Performance with MicroStrat...
 
Augmented OLAP for Big Data Analytics
Augmented OLAP for Big Data AnalyticsAugmented OLAP for Big Data Analytics
Augmented OLAP for Big Data Analytics
 
Addressing the systemic shortcomings of cloud analytics
Addressing the systemic shortcomings of cloud analyticsAddressing the systemic shortcomings of cloud analytics
Addressing the systemic shortcomings of cloud analytics
 
Apache Kylin and Use Cases - 2018 Big Data Spain
Apache Kylin and Use Cases - 2018 Big Data SpainApache Kylin and Use Cases - 2018 Big Data Spain
Apache Kylin and Use Cases - 2018 Big Data Spain
 
Ian Uriarte Timbergrove at IBM IoTExchange 2019
Ian Uriarte Timbergrove at IBM IoTExchange 2019Ian Uriarte Timbergrove at IBM IoTExchange 2019
Ian Uriarte Timbergrove at IBM IoTExchange 2019
 
Cisco Connect Toronto 2018 an introduction to Cisco kinetic
Cisco Connect Toronto 2018   an introduction to Cisco kineticCisco Connect Toronto 2018   an introduction to Cisco kinetic
Cisco Connect Toronto 2018 an introduction to Cisco kinetic
 
Cisco Connect Toronto 2018 an introduction to Cisco kinetic
Cisco Connect Toronto 2018   an introduction to Cisco kineticCisco Connect Toronto 2018   an introduction to Cisco kinetic
Cisco Connect Toronto 2018 an introduction to Cisco kinetic
 
IoT Update | Hoe implementeer je IoT Schaalbaar in je IT landschap
IoT Update | Hoe implementeer je IoT Schaalbaar in je IT landschapIoT Update | Hoe implementeer je IoT Schaalbaar in je IT landschap
IoT Update | Hoe implementeer je IoT Schaalbaar in je IT landschap
 
IBM CDS Overview
IBM CDS OverviewIBM CDS Overview
IBM CDS Overview
 
The value of a connected factory
The value of a connected factoryThe value of a connected factory
The value of a connected factory
 
A Connected Data Landscape: Virtualization and the Internet of Things
A Connected Data Landscape: Virtualization and the Internet of ThingsA Connected Data Landscape: Virtualization and the Internet of Things
A Connected Data Landscape: Virtualization and the Internet of Things
 
Customer Presentation - IBM Cloud Pak for Data Overview (Level 100).PPTX
Customer Presentation - IBM Cloud Pak for Data Overview (Level 100).PPTXCustomer Presentation - IBM Cloud Pak for Data Overview (Level 100).PPTX
Customer Presentation - IBM Cloud Pak for Data Overview (Level 100).PPTX
 

Plus de Tyler Wishnoff

Hassle-Free Data Lake Governance: Automating Your Analytics with a Semantic L...
Hassle-Free Data Lake Governance: Automating Your Analytics with a Semantic L...Hassle-Free Data Lake Governance: Automating Your Analytics with a Semantic L...
Hassle-Free Data Lake Governance: Automating Your Analytics with a Semantic L...Tyler Wishnoff
 
Providing Interactive Analytics on Excel with Billions of Rows
Providing Interactive Analytics on Excel with Billions of RowsProviding Interactive Analytics on Excel with Billions of Rows
Providing Interactive Analytics on Excel with Billions of RowsTyler Wishnoff
 
Apache kylin 101 - Get Sub-Second Analytics on Massive Datasets
Apache kylin 101 - Get Sub-Second Analytics on Massive DatasetsApache kylin 101 - Get Sub-Second Analytics on Massive Datasets
Apache kylin 101 - Get Sub-Second Analytics on Massive DatasetsTyler Wishnoff
 
AI-Powered Analytics: What It Is and How It’s Powering the Next Generation of...
AI-Powered Analytics: What It Is and How It’s Powering the Next Generation of...AI-Powered Analytics: What It Is and How It’s Powering the Next Generation of...
AI-Powered Analytics: What It Is and How It’s Powering the Next Generation of...Tyler Wishnoff
 
Analysis of the Pressure Placed on Medical Systems during the COVID-19 Pandemic
Analysis of the Pressure Placed on Medical Systems during the COVID-19 PandemicAnalysis of the Pressure Placed on Medical Systems during the COVID-19 Pandemic
Analysis of the Pressure Placed on Medical Systems during the COVID-19 PandemicTyler Wishnoff
 
Apache Kylin Meetup: Berlin - With OLX Group
Apache Kylin Meetup: Berlin - With OLX GroupApache Kylin Meetup: Berlin - With OLX Group
Apache Kylin Meetup: Berlin - With OLX GroupTyler Wishnoff
 
Apache Kylin Data Summit 2019: Kyligence Presentation
Apache Kylin Data Summit 2019: Kyligence PresentationApache Kylin Data Summit 2019: Kyligence Presentation
Apache Kylin Data Summit 2019: Kyligence PresentationTyler Wishnoff
 
How Analytics Teams Using SSAS Can Embrace Big Data and the Cloud
How Analytics Teams Using SSAS Can Embrace Big Data and the CloudHow Analytics Teams Using SSAS Can Embrace Big Data and the Cloud
How Analytics Teams Using SSAS Can Embrace Big Data and the CloudTyler Wishnoff
 
Accelerating Big Data Analytics with Apache Kylin
Accelerating Big Data Analytics with Apache KylinAccelerating Big Data Analytics with Apache Kylin
Accelerating Big Data Analytics with Apache KylinTyler Wishnoff
 

Plus de Tyler Wishnoff (9)

Hassle-Free Data Lake Governance: Automating Your Analytics with a Semantic L...
Hassle-Free Data Lake Governance: Automating Your Analytics with a Semantic L...Hassle-Free Data Lake Governance: Automating Your Analytics with a Semantic L...
Hassle-Free Data Lake Governance: Automating Your Analytics with a Semantic L...
 
Providing Interactive Analytics on Excel with Billions of Rows
Providing Interactive Analytics on Excel with Billions of RowsProviding Interactive Analytics on Excel with Billions of Rows
Providing Interactive Analytics on Excel with Billions of Rows
 
Apache kylin 101 - Get Sub-Second Analytics on Massive Datasets
Apache kylin 101 - Get Sub-Second Analytics on Massive DatasetsApache kylin 101 - Get Sub-Second Analytics on Massive Datasets
Apache kylin 101 - Get Sub-Second Analytics on Massive Datasets
 
AI-Powered Analytics: What It Is and How It’s Powering the Next Generation of...
AI-Powered Analytics: What It Is and How It’s Powering the Next Generation of...AI-Powered Analytics: What It Is and How It’s Powering the Next Generation of...
AI-Powered Analytics: What It Is and How It’s Powering the Next Generation of...
 
Analysis of the Pressure Placed on Medical Systems during the COVID-19 Pandemic
Analysis of the Pressure Placed on Medical Systems during the COVID-19 PandemicAnalysis of the Pressure Placed on Medical Systems during the COVID-19 Pandemic
Analysis of the Pressure Placed on Medical Systems during the COVID-19 Pandemic
 
Apache Kylin Meetup: Berlin - With OLX Group
Apache Kylin Meetup: Berlin - With OLX GroupApache Kylin Meetup: Berlin - With OLX Group
Apache Kylin Meetup: Berlin - With OLX Group
 
Apache Kylin Data Summit 2019: Kyligence Presentation
Apache Kylin Data Summit 2019: Kyligence PresentationApache Kylin Data Summit 2019: Kyligence Presentation
Apache Kylin Data Summit 2019: Kyligence Presentation
 
How Analytics Teams Using SSAS Can Embrace Big Data and the Cloud
How Analytics Teams Using SSAS Can Embrace Big Data and the CloudHow Analytics Teams Using SSAS Can Embrace Big Data and the Cloud
How Analytics Teams Using SSAS Can Embrace Big Data and the Cloud
 
Accelerating Big Data Analytics with Apache Kylin
Accelerating Big Data Analytics with Apache KylinAccelerating Big Data Analytics with Apache Kylin
Accelerating Big Data Analytics with Apache Kylin
 

Dernier

NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...Boston Institute of Analytics
 
IMA MSN - Medical Students Network (2).pptx
IMA MSN - Medical Students Network (2).pptxIMA MSN - Medical Students Network (2).pptx
IMA MSN - Medical Students Network (2).pptxdolaknnilon
 
DBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfDBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfJohn Sterrett
 
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...Florian Roscheck
 
Building on a FAIRly Strong Foundation to Connect Academic Research to Transl...
Building on a FAIRly Strong Foundation to Connect Academic Research to Transl...Building on a FAIRly Strong Foundation to Connect Academic Research to Transl...
Building on a FAIRly Strong Foundation to Connect Academic Research to Transl...Jack DiGiovanna
 
Multiple time frame trading analysis -brianshannon.pdf
Multiple time frame trading analysis -brianshannon.pdfMultiple time frame trading analysis -brianshannon.pdf
Multiple time frame trading analysis -brianshannon.pdfchwongval
 
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024thyngster
 
Identifying Appropriate Test Statistics Involving Population Mean
Identifying Appropriate Test Statistics Involving Population MeanIdentifying Appropriate Test Statistics Involving Population Mean
Identifying Appropriate Test Statistics Involving Population MeanMYRABACSAFRA2
 
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhi
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝DelhiRS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhi
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhijennyeacort
 
Top 5 Best Data Analytics Courses In Queens
Top 5 Best Data Analytics Courses In QueensTop 5 Best Data Analytics Courses In Queens
Top 5 Best Data Analytics Courses In Queensdataanalyticsqueen03
 
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM TRACKING WITH GOOGLE ANALYTICS.pptx
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM  TRACKING WITH GOOGLE ANALYTICS.pptxEMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM  TRACKING WITH GOOGLE ANALYTICS.pptx
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM TRACKING WITH GOOGLE ANALYTICS.pptxthyngster
 
RadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfRadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfgstagge
 
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...soniya singh
 
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)jennyeacort
 
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...dajasot375
 
PKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptxPKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptxPramod Kumar Srivastava
 
20240419 - Measurecamp Amsterdam - SAM.pdf
20240419 - Measurecamp Amsterdam - SAM.pdf20240419 - Measurecamp Amsterdam - SAM.pdf
20240419 - Measurecamp Amsterdam - SAM.pdfHuman37
 
MK KOMUNIKASI DATA (TI)komdat komdat.docx
MK KOMUNIKASI DATA (TI)komdat komdat.docxMK KOMUNIKASI DATA (TI)komdat komdat.docx
MK KOMUNIKASI DATA (TI)komdat komdat.docxUnduhUnggah1
 

Dernier (20)

NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
 
IMA MSN - Medical Students Network (2).pptx
IMA MSN - Medical Students Network (2).pptxIMA MSN - Medical Students Network (2).pptx
IMA MSN - Medical Students Network (2).pptx
 
DBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfDBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdf
 
E-Commerce Order PredictionShraddha Kamble.pptx
E-Commerce Order PredictionShraddha Kamble.pptxE-Commerce Order PredictionShraddha Kamble.pptx
E-Commerce Order PredictionShraddha Kamble.pptx
 
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
 
Deep Generative Learning for All - The Gen AI Hype (Spring 2024)
Deep Generative Learning for All - The Gen AI Hype (Spring 2024)Deep Generative Learning for All - The Gen AI Hype (Spring 2024)
Deep Generative Learning for All - The Gen AI Hype (Spring 2024)
 
Building on a FAIRly Strong Foundation to Connect Academic Research to Transl...
Building on a FAIRly Strong Foundation to Connect Academic Research to Transl...Building on a FAIRly Strong Foundation to Connect Academic Research to Transl...
Building on a FAIRly Strong Foundation to Connect Academic Research to Transl...
 
Multiple time frame trading analysis -brianshannon.pdf
Multiple time frame trading analysis -brianshannon.pdfMultiple time frame trading analysis -brianshannon.pdf
Multiple time frame trading analysis -brianshannon.pdf
 
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
 
Identifying Appropriate Test Statistics Involving Population Mean
Identifying Appropriate Test Statistics Involving Population MeanIdentifying Appropriate Test Statistics Involving Population Mean
Identifying Appropriate Test Statistics Involving Population Mean
 
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhi
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝DelhiRS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhi
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhi
 
Top 5 Best Data Analytics Courses In Queens
Top 5 Best Data Analytics Courses In QueensTop 5 Best Data Analytics Courses In Queens
Top 5 Best Data Analytics Courses In Queens
 
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM TRACKING WITH GOOGLE ANALYTICS.pptx
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM  TRACKING WITH GOOGLE ANALYTICS.pptxEMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM  TRACKING WITH GOOGLE ANALYTICS.pptx
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM TRACKING WITH GOOGLE ANALYTICS.pptx
 
RadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfRadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdf
 
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
 
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
 
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...
 
PKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptxPKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptx
 
20240419 - Measurecamp Amsterdam - SAM.pdf
20240419 - Measurecamp Amsterdam - SAM.pdf20240419 - Measurecamp Amsterdam - SAM.pdf
20240419 - Measurecamp Amsterdam - SAM.pdf
 
MK KOMUNIKASI DATA (TI)komdat komdat.docx
MK KOMUNIKASI DATA (TI)komdat komdat.docxMK KOMUNIKASI DATA (TI)komdat komdat.docx
MK KOMUNIKASI DATA (TI)komdat komdat.docx
 

How to Guarantee Exact COUNT DISTINCT Queries with Sub-Second Latency on Massive Datasets

  • 1. How to Guarantee Exact COUNT DISTINCT Queries with Sub-Second Latency on Massive Datasets Kaige Liu 2020.5
  • 2. © Kyligence Inc. 2019, Confidential. Business Scenarios Technical Principles Demo Use Cases Q&A Agenda
  • 3. © Kyligence Inc. 2019, Confidential. Business Scenarios
  • 4. © Kyligence Inc. 2019, Confidential. What Is Count Distinct? Count Distinct is used to compute the number of unique values in a data set. • PV (Page View) • UV (Unique Visitors) ID Username Page 1 Alice /kyligence 2 Alice /Kyligence/Blog 3 Carol /Kyligence/Events 4 Bob /Kyligence/Resources 5 Alice /Kyligence/Downloads Alice, Bob, Carol 3
  • 5. © Kyligence Inc. 2019, Confidential. Approximate and Exact Count Distinct • Approximate Count Distinct • Quick, less memory/CPU • Not accurate • Trend analysis, small errors are acceptable • Exact Count Distinct • Slow, more memory/CPU • Accurate • Transaction relevant. Paid Advertising, Precision Marketing, etc. Error Rate $ 1 Million $ 1 Billion 1.22% $12,200 $12,200,000 2.44% $24,000 $24,000,000 9.75% $97,500 $97,500,000
  • 6. © Kyligence Inc. 2019, Confidential. Where are they coming from? Who are my visitors? Web/Ap p Analytic s Which page lost the most users? How many active users? How many new users? How many unique visitors? Scenarios - Web/App Analytics
  • 7. © Kyligence Inc. 2019, Confidential. Scenarios - User Behavior Analytics Retention Analysis Funnel Analysis
  • 8. © Kyligence Inc. 2019, Confidential. Technical Principles
  • 9. © Kyligence Inc. 2019, Confidential. Challenges with Exact Count Distinct • Approximate Count Distinct is easy – HyperLogLog • Exact Count Distinct is a big challenge for all query engines at massive scale Challenges • Bad performance – Need to scan all data • Non-cumulative – Hard to do rollup and/or operations • Hard to optimize on multiple columns • Analysis always requires more than one count distinct operation
  • 10. © Kyligence Inc. 2019, Confidential. Count Distinct Performance on Different Platforms • Google BigQuery • Snowflake • Athena • Apache Kylin • Kyligence
  • 11. © Kyligence Inc. 2019, Confidential. Kyligence = Kylin + Intelligence • Founded in 2016 by the creators of Apache Kylin • Built around Kylin, with augmented AI and enhanced to deliver unprecedented enterprise analytic performance • CRN Top-10 big data startups in 2018 • Global Presence: San Jose, Seattle, New York, Shanghai, Beijing • VCs: Fidelity International, Shunwei Capital, Broadband Capital, Redpoint, Cisco, Coatue Accelerate Critical Business Decisions with AI-Augmented Data Management and Analytics 2016 Founded Pre- A Redpoint Cisco 2017 Series A CBC Shunwei 2018 Series B 8Roads 2019 Series C Coatue
  • 12. © Kyligence Inc. 2019, Confidential. How Does Apache Kylin Achieve This? BitmapPre-Aggregation • Pre-aggregate count distinct in cubes • Fetch results directly without on the fly calculations • Supports Rollup • Reduces memory/storage significantly • Supports String type and detail queries Dictionary
  • 13. © Kyligence Inc. 2019, Confidential. Pre-Aggregation Date UID Page 2020-04-01 01 1 /kyligence 2020-04-01 01 1 /Kyligence/Blog 2020-04-01 01 2 /Kyligence/News 2020-04-02 02 3 /Kyligence/Events 2020-04-02 02 2 /Kyligence/Resources 2020-04-02 02 1 /Kyligence/Downloads Date Count(UID) Count(distinct UID) UID) 2020-04-01 01 3 2 2020-04-02 02 3 3 Date Count(UID) Count(distinct UID) 2020-04-01 01 and 2020-04-02 02 6 ??
  • 14. © Kyligence Inc. 2019, Confidential. 7 6 5 4 3 2 1 0 Bitmap UID 1 2 4 5 7 9 10 11 13 1 0 0 1 0 1 1 0 0 0 1 0 1 1 1 0 Table Bitmap • Saves storage significantly • Supports logical operations directly • Contains information needed to do aggregation • RoaringBitmap
  • 15. © Kyligence Inc. 2019, Confidential. Bitmap Date UID Page 2020-04-01 01 1 /kyligence 2020-04-01 01 1 /Kyligence/Blog 2020-04-01 01 2 /Kyligence/News 2020-04-02 02 3 /Kyligence/Events 2020-04-02 02 2 /Kyligence/Resources 2020-04-02 02 1 /Kyligence/Downloads Date Count(UID) Count(distinct UID) UID) 2020-04-01 01 3 2 2020-04-02 02 3 3 Date Count(UID) Count(distinct UID) 2020-04-01 01 and 2020-04-02 02 6 Bitmap(1,2,3) Date Count(UID) Count(distinct UID) UID) 2020-04-01 3 Bitmap(1,2) 2020-04-02 3 Bitmap(1,2,3)
  • 16. © Kyligence Inc. 2019, Confidential. Operations in Bitmap • Two bitmaps, each containing two different data sets: [1, 3, 4, 5] [2, 3, 4, 6] • And - All elements contained in both bitmaps: [1, 3, 4, 5] and [2, 3, 4, 6] = [3, 4] Scenarios: Retention Analysis, Funnel Analysis • Or – All elements in either bitmap: [1, 3, 4, 5] or [2, 3, 4, 6] = [1, 2, 3, 4, 5, 6] Scenarios: Cross-Dimension Analysis
  • 17. © Kyligence Inc. 2019, Confidential. Dictionary Date USERNAME Page 2020-04-01 01 Alice /kyligence 2020-04-01 01 Alice /Kyligence/Blog 2020-04-01 01 Bob /Kyligence/News 2020-04-02 02 Coral /Kyligence/Events 2020-04-02 02 Bob /Kyligence/Resources 2020-04-02 02 Alice /Kyligence/Downloads USERNAME ECODED Alice 1 Bob 2 Coral 3 Date Count(UID) Count(distinct UID) 2020-04-01 01 and 2020-04-02 02 6 Bitmap(1,2,3) Date Count(UID) Count(distinct UID) UID) 2020-04-01 01 3 Bitmap(1,2) 2020-04-02 02 3 Bitmap(1,2,3) Bitmap can only support int values. How about String columns? Dictionary
  • 18. © Kyligence Inc. 2019, Confidential. Use Cases
  • 19. © Kyligence Inc. 2019, Confidential. Manbang Group • The largest Chinese truck logistics startup • 7 million+ trucks • 2.25 million active users • 8 apps and 10 TB+ data Requirements • Retention analysis on a wide range of dimensions and date ranges • Funnel analysis with ability to customize funnel • User profile analysis
  • 20. © Kyligence Inc. 2019, Confidential. Architecture with Apache Kylin
  • 21. © Kyligence Inc. 2019, Confidential. Retention Analysis for Manbang Group • Users can choose any column and any date range to do the retention analysis
  • 22. © Kyligence Inc. 2019, Confidential. Funnel Analysis for Manbang group • Users can customize funnels with any number of steps • Can identify the specific users lost between steps
  • 23. © Kyligence Inc. 2019, Confidential. DiDi • #1 ride-share company in China • 92 million monthly active users (as of Dec. 2019) • 24 million rides per day in 2019 Requirements • User profile analysis • Precision marketing
  • 24. © Kyligence Inc. 2019, Confidential. Scenarios – Apache Kylin in Didi • Precision Marketing o Send coupons to exact target users o Upgrade cars for specific users • Promotion Activity Analysis o How many new/returned users are gained in this activity? o Which kind of users are most interested in this activity? • Optimize User Experience o Which stages lost the most users? o How to increase customer stickiness? User Profile Precision Marketing User Behavior Analysis User Tags Workflow Analysis Promotion Activity Analysis
  • 25. © Kyligence Inc. 2019, Confidential. Didi Kylin Usage 200 TB+ 5,000+ 7,000+ 7 Data Cubes Jobs per day Clusters
  • 26. © Kyligence Inc. 2019, Confidential. Join the Community https://github.com/apache/kylin apache-kylin.slack.comuser@kylin.apache.org

Notes de l'éditeur

  1. UV/PV put some words in the slide
  2. Put a static image instead of gif
  3. Link And OR to analysis scenarios