Ce diaporama a bien été signalé.
Nous utilisons votre profil LinkedIn et vos données d’activité pour vous proposer des publicités personnalisées et pertinentes. Vous pouvez changer vos préférences de publicités à tout moment.
FLINK-POWERED
CUSTOMER EXPERIENCE:
SCALING FROM 5 BILLION
DOWN TO ONE
Dave Torok
Distinguished Architect
Comcast Corporati...
3
COMCAST CUSTOMER RELATIONSHIPS
30.3 MILLION OVERALL CUSTOMER
RELATIONSHIPS AT 2018 YEAR END
25.1 MILLION RESIDENTIAL HIG...
4
DELIVER THE ULTIMATE CUSTOMER EXPERIENCE
IS THE CUSTOMER HAVING A GOOD EXPERIENCE
FOR HIGH SPEED DATA (HSD) SERVICE?
IF ...
5
REDUCE COST AND TIME TO RESOLUTION
CUSTOMER
SELF-SERVICE
COST
$ ~0
TIME
< 15 MINUTES
TALK TO AN
AGENT
COST ORDER OF
MAGN...
How do we personalize
the conversation?
7
CUSTOMER EXPERIENCE INDICATORS
Comcast collects, stores, and uses all data in accordance with our privacy disclosures to...
8
HIGH LEVEL CONCEPT
Customer
Experience
Diagnosis
Platform
Service
Delivery
Systems
Predictions /
Diagnosis /
Treatments
...
9
SIMPLIFY INDICATORS
"packetErrors": [“327"],
"systemUptime":
{"string": "4:26:34.27"},
"octets": ["174756692"],
"timerEr...
1 0
SIMPLIFY INDICATORS
"packetErrors": [“327"],
"systemUptime":
{"string": "4:26:34.27"},
"octets": ["174756692"],
"timer...
1 1
SIMPLIFY INDICATORS
"packetErrors": [“327"],
"systemUptime":
{"string": "4:26:34.27"},
"octets": ["174756692"],
"timer...
What is the
Digital Experience?
1 3
XFINITY ASSISTANT EXPERIENCE
1 4
XFINITY ASSISTANT EXPERIENCE
My internet is slow My internet is slow
1 5
XFINITY ASSISTANT EXPERIENCE
FLINK
POWER!
1 6
XFINITY ASSISTANT EXPERIENCE
1 7
XFINITY ASSISTANT EXPERIENCE
1 8
XFINITY ASSISTANT EXPERIENCE
1 9
XFINITY ASSISTANT EXPERIENCE
2 0
XFINITY ASSISTANT EXPERIENCE
FOLLOWING UP ON THE
INTERACTION:
Is the problem resolved?
If so, great!
If not, offer to ...
2 1
POWERED BY APACHE FLINK
®
Online Feature
Store
Model/ Feature
Metadata
Feature
Assembly
Feature Sink
Model
Execution
P...
2 2
CHALLENGES ALONG THE WAY
THE
TRIGGER AND
DIAGNOSIS
PROBLEM
THE
FEATURE STORE
PROBLEM
THE
CHECKPOINT
PROBLEM
THE
REST
P...
2 3
DATA POINT VOLUME MAY–NOVEMBER 2018
2 4
DATA POINT VOLUME MAY–NOVEMBER 2018
OCTOBER 2018 – 1 BILLION
2 5
DATA POINT VOLUME MAY–NOVEMBER 2018
OCTOBER 2018 – 1 BILLION
NOVEMBER 2018 – 2 BILLION
2 6
LET’S ADD ANOTHER 3 BILLION!
15 MILLION DEVICES
3 BILLION DATA POINTS
2 7
MEETING THE CHALLENGE
HORIZONTAL
SCALING
TO MEET THE VOLUME
VERTICAL
SCALING
TO PROCESS HIGH-
VOLUME DATA SOURCES
FLIN...
2 8
USING FLINK STATE TO TRIGGER DIAGNOSIS
GREEN
GREEN
GREEN
RED
GREEN
GREEN
2 9
USING FLINK STATE TO TRIGGER DIAGNOSIS
GREEN
GREEN
GREEN
RED
GREEN
GREEN
GREEN
RED
GREEN
RED
3 0
USING FLINK STATE TO TRIGGER DIAGNOSIS
GREEN
GREEN
GREEN
RED
GREEN
GREEN
GREEN
RED
GREEN
REDDETECT STATE CHANGE
3 1
USING FLINK STATE TO TRIGGER DIAGNOSIS
GREEN
GREEN
GREEN
RED
GREEN
GREEN
GREEN
GREEN
GREEN
RED
GREEN
RED
GREEN
RED
GRE...
3 2
KEEPING TRACK OF DATA POINT STATE
QUERYABLE STATE
FEATURE STORE
UP TO 15KB JSON
PER DATA POINT
~ 1 TB TOTAL
RED / GREE...
3 3
WHERE ARE WE TODAY
INDICATORS HIGH-VOLUME
INDICATORS TOTAL
725+
MILLION
DATA
POINTS
PER DAY
6 BILLION
PER DAY
~7 BILLI...
3 4
WHERE ARE WE TODAY – FLINK CLUSTERS
CLUSTERS
14
INSTANCES
150
VCPU
1100
RAM
5.8 TB
3 5
RESULTS
MILLIONS
OF UNIQUE USERS
REACHED
MILLIONS
OF OPERATIONAL
DOLLARS SAVED
INCREASED
CUSTOMER
SATISFACTION WITH
DI...
3 6
RESULTS
MILLIONS
OF UNIQUE USERS
REACHED
MILLIONS
OF OPERATIONAL
DOLLARS SAVED
INCREASED
CUSTOMER
SATISFACTION WITH
DI...
ADVENTURES IN SCALING
FROM ZERO TO 5 7 BILLION
DATA POINTS PER DAY
5:30 PM – 6:10 PM
Carmel Meeting Room
For more details,...
3 8
WE’RE
HIRING!
PHILADELPHIA
WASHINGTON, D.C.
MOUNTAIN VIEW
DENVER
Flink Forward San Francisco 2019: Flink Powered Customer Experience: Scaling from 5 Billion down to One - Dave Torok
Flink Forward San Francisco 2019: Flink Powered Customer Experience: Scaling from 5 Billion down to One - Dave Torok
Prochain SlideShare
Chargement dans…5
×

Flink Forward San Francisco 2019: Flink Powered Customer Experience: Scaling from 5 Billion down to One - Dave Torok

322 vues

Publié le

Flink Powered Customer Experience: Scaling from 5 Billion down to One
What turns a simple interaction into a great customer experience? How do we transform a digital interaction into a personalized conversation? Over the past year our team scaled a Customer Experience use case to process over 5 Billion data points per day using Apache Flink. This presentation will show how Flink helps deliver a personalized, contextual interaction by scaling the solution down to one…the customer.

Publié dans : Technologie
  • Soyez le premier à commenter

  • Soyez le premier à aimer ceci

Flink Forward San Francisco 2019: Flink Powered Customer Experience: Scaling from 5 Billion down to One - Dave Torok

  1. 1. FLINK-POWERED CUSTOMER EXPERIENCE: SCALING FROM 5 BILLION DOWN TO ONE Dave Torok Distinguished Architect Comcast Corporation 2 April, 2019 Flink Forward – San Francisco 2019
  2. 2. 3 COMCAST CUSTOMER RELATIONSHIPS 30.3 MILLION OVERALL CUSTOMER RELATIONSHIPS AT 2018 YEAR END 25.1 MILLION RESIDENTIAL HIGH-SPEED INTERNET CUSTOMERS AT 2018 YEAR END 1.2 MILLION RESIDENTIAL HIGH-SPEED INTERNET CUSTOMER NET ADDITIONS IN 2018
  3. 3. 4 DELIVER THE ULTIMATE CUSTOMER EXPERIENCE IS THE CUSTOMER HAVING A GOOD EXPERIENCE FOR HIGH SPEED DATA (HSD) SERVICE? IF THE CUSTOMER ENGAGES US DIGITALLY, CAN WE OFFER A SELF-SERVICE SOLUTION? IF THERE IS AN ISSUE CAN WE OFFER OUR AGENTS AND TECHNICIANS A DIAGNOSIS TO HELP SOLVE THE PROBLEM QUICKER? REDUCE THE TIME TO RESOLVE ISSUES REDUCE COST TO THE BUSINESS AND THE CUSTOMER
  4. 4. 5 REDUCE COST AND TIME TO RESOLUTION CUSTOMER SELF-SERVICE COST $ ~0 TIME < 15 MINUTES TALK TO AN AGENT COST ORDER OF MAGNITUDE: $ 5-10 TIME: < AN HOUR TECHNICIAN VISIT TO HOME COST ORDER OF MAGNITUDE: $ 50-100 TIME: AT LEAST A DAY
  5. 5. How do we personalize the conversation?
  6. 6. 7 CUSTOMER EXPERIENCE INDICATORS Comcast collects, stores, and uses all data in accordance with our privacy disclosures to users and applicable laws.
  7. 7. 8 HIGH LEVEL CONCEPT Customer Experience Diagnosis Platform Service Delivery Systems Predictions / Diagnosis / Treatments CUSTOMER ENGAGEMENT PLATFORM CUSTOMER EXPERIENCE INDICATORS Comcast collects, stores, and uses all data in accordance with our privacy disclosures to users and applicable laws.
  8. 8. 9 SIMPLIFY INDICATORS "packetErrors": [“327"], "systemUptime": {"string": "4:26:34.27"}, "octets": ["174756692"], "timerError": { "int": 0 }, "sysInfoHwphase2": {"int": 1} "firmwareDlError": { int": 0 }, "sysErrorTr69NotRegistered": null, "sysShSnmpsubagentRestartMw": null, "wifiErrorKernelDriverNotconnected": null, "forcedNonDeviceRebootError": { "int": 0 }
  9. 9. 1 0 SIMPLIFY INDICATORS "packetErrors": [“327"], "systemUptime": {"string": "4:26:34.27"}, "octets": ["174756692"], "timerError": { "int": 0 }, "sysInfoHwphase2": {"int": 1} "firmwareDlError": { int": 0 }, "sysErrorTr69NotRegistered": null, "sysShSnmpsubagentRestartMw": null, "wifiErrorKernelDriverNotconnected": null, "forcedNonDeviceRebootError": { "int": 0 } GREEN CUSTOMER EXPERIENCE IS GOOD
  10. 10. 1 1 SIMPLIFY INDICATORS "packetErrors": [“327"], "systemUptime": {"string": "4:26:34.27"}, "octets": ["174756692"], "timerError": { "int": 0 }, "sysInfoHwphase2": {"int": 1} "firmwareDlError": { int": 0 }, "sysErrorTr69NotRegistered": null, "sysShSnmpsubagentRestartMw": null, "wifiErrorKernelDriverNotconnected": null, "forcedNonDeviceRebootError": { "int": 0 } GREEN CUSTOMER EXPERIENCE IS GOOD RED POSSIBLE IMPACT TO CUSTOMER EXPERIENCE
  11. 11. What is the Digital Experience?
  12. 12. 1 3 XFINITY ASSISTANT EXPERIENCE
  13. 13. 1 4 XFINITY ASSISTANT EXPERIENCE My internet is slow My internet is slow
  14. 14. 1 5 XFINITY ASSISTANT EXPERIENCE FLINK POWER!
  15. 15. 1 6 XFINITY ASSISTANT EXPERIENCE
  16. 16. 1 7 XFINITY ASSISTANT EXPERIENCE
  17. 17. 1 8 XFINITY ASSISTANT EXPERIENCE
  18. 18. 1 9 XFINITY ASSISTANT EXPERIENCE
  19. 19. 2 0 XFINITY ASSISTANT EXPERIENCE FOLLOWING UP ON THE INTERACTION: Is the problem resolved? If so, great! If not, offer to talk with an agent.
  20. 20. 2 1 POWERED BY APACHE FLINK ® Online Feature Store Model/ Feature Metadata Feature Assembly Feature Sink Model Execution Prediction / Outcome Store Customer Experience Data Point Producers FLINK LOCAL STATE State Flow And Trigger FLINK LOCAL STATE FLINK LOCAL STATE Feature Store Manager FLINK LOCAL STATE Historical Feature Store Prediction Store Sink Apache®, Apache Flink®, and the squirrel logo are either registered trademarks or trademarks of the Apache Software Foundation in the United States and/or other countries.
  21. 21. 2 2 CHALLENGES ALONG THE WAY THE TRIGGER AND DIAGNOSIS PROBLEM THE FEATURE STORE PROBLEM THE CHECKPOINT PROBLEM THE REST PROBLEM THE VOLUME PROBLEM THE TRIGGER AND DIAGNOSIS PROBLEM THE INEFFICIENT OBJECT HANDLING PROBLEM THE CUSTOMER STATE PROBLEM THE REALLY HIGH VOLUME AND FEATURE STORE PROBLEM #2
  22. 22. 2 3 DATA POINT VOLUME MAY–NOVEMBER 2018
  23. 23. 2 4 DATA POINT VOLUME MAY–NOVEMBER 2018 OCTOBER 2018 – 1 BILLION
  24. 24. 2 5 DATA POINT VOLUME MAY–NOVEMBER 2018 OCTOBER 2018 – 1 BILLION NOVEMBER 2018 – 2 BILLION
  25. 25. 2 6 LET’S ADD ANOTHER 3 BILLION! 15 MILLION DEVICES 3 BILLION DATA POINTS
  26. 26. 2 7 MEETING THE CHALLENGE HORIZONTAL SCALING TO MEET THE VOLUME VERTICAL SCALING TO PROCESS HIGH- VOLUME DATA SOURCES FLINK FEATURES QUERYABLE STATE CHECKPOINTING GLOBAL WINDOWS CONNECTED STREAMS STREAMING SINKS
  27. 27. 2 8 USING FLINK STATE TO TRIGGER DIAGNOSIS GREEN GREEN GREEN RED GREEN GREEN
  28. 28. 2 9 USING FLINK STATE TO TRIGGER DIAGNOSIS GREEN GREEN GREEN RED GREEN GREEN GREEN RED GREEN RED
  29. 29. 3 0 USING FLINK STATE TO TRIGGER DIAGNOSIS GREEN GREEN GREEN RED GREEN GREEN GREEN RED GREEN REDDETECT STATE CHANGE
  30. 30. 3 1 USING FLINK STATE TO TRIGGER DIAGNOSIS GREEN GREEN GREEN RED GREEN GREEN GREEN GREEN GREEN RED GREEN RED GREEN RED GREEN REDDETECT STATE CHANGE SEND FOR DIAGNOSIS
  31. 31. 3 2 KEEPING TRACK OF DATA POINT STATE QUERYABLE STATE FEATURE STORE UP TO 15KB JSON PER DATA POINT ~ 1 TB TOTAL RED / GREEN “TRIGGER” [ENTITY ID] (DATA POINT TYPE, STATE) ~1 GB TOTAL 25+ MILLION HIGH SPEED INTERNET CUSTOMERS 10+ DATA POINT TYPES TWO FLINK STATES TO MANAGE:
  32. 32. 3 3 WHERE ARE WE TODAY INDICATORS HIGH-VOLUME INDICATORS TOTAL 725+ MILLION DATA POINTS PER DAY 6 BILLION PER DAY ~7 BILLION DATA POINTS PER DAY
  33. 33. 3 4 WHERE ARE WE TODAY – FLINK CLUSTERS CLUSTERS 14 INSTANCES 150 VCPU 1100 RAM 5.8 TB
  34. 34. 3 5 RESULTS MILLIONS OF UNIQUE USERS REACHED MILLIONS OF OPERATIONAL DOLLARS SAVED INCREASED CUSTOMER SATISFACTION WITH DIGITAL INTERACTION REDUCED TIME TO RESOLVE COMMON ISSUES THROUGH SELF- SERVICE
  35. 35. 3 6 RESULTS MILLIONS OF UNIQUE USERS REACHED MILLIONS OF OPERATIONAL DOLLARS SAVED INCREASED CUSTOMER SATISFACTION WITH DIGITAL INTERACTION REDUCED TIME TO RESOLVE COMMON ISSUES THROUGH SELF- SERVICE SCALING THE EXPERIENCE TO NE
  36. 36. ADVENTURES IN SCALING FROM ZERO TO 5 7 BILLION DATA POINTS PER DAY 5:30 PM – 6:10 PM Carmel Meeting Room For more details, please attend my technical presentation!
  37. 37. 3 8 WE’RE HIRING! PHILADELPHIA WASHINGTON, D.C. MOUNTAIN VIEW DENVER

×