advancing the traditional enterprise: an ea story
TRANSCRIPT
ELECTRONIC ARTS COMPANY PROFILE
Founded: 1982
#1 Market Share in NA and Europe
FY12 Revenue: US $4.1B
Platform Partnerships: Sony, Microsoft,
Nintendo, Apple, Facebook, Google,
Amazon
Exclusive Partnerships: FIFA, PGA
TOUR, ESPN, NFL, Hasbro
Distribution: >70 Countries
World’s Largest Studio Operation
Major Studios: in Canada, US, UK,
Sweden
VIDEO GAMES NOW – DIGITAL PLAYGROUND
GAMES NOW – DIGITAL PLAYGROUND
Virtual Reality
Created with Player Content
Content Rich
THE GAME (R)EVOLUTION
Pay upfront Pay over time
Play alone Play with friends
Box products Digital services
Channel distribution Direct distribution
Agility / Availability / Throughput / Flexibility
Cost / Sustainability / Reusability
Processing
Event / Mini-Batch / Batch
Infrastructure Capture
Servers
Delivery
Channels
Storage
PIECES OF THE ANALYTICS PUZZLE
We’ve always had the Volume, but Velocity and Variety demands have quadrupled.
Volume Velocity Variety
Big Data
BACKGROUND INFORMATION
• Robust workhorse hardware
• SMP Database engine
• Non-RI DB models (RI checks at ETL layer)
• Multi-threaded process
• Primitive workload management
• Mostly direct DB connect data capture
• Extensive log parsing / processing
• Heavy lifting at ETL layer
• Vertical scalability
• Strong network backbone
TECH OPTION 1
Managing load
window
TECH OPTION 1 - CHALLENGES
Heavy upfront investment
in hardware
Capacity expansion does
not yield linear growth
Primitive mixed
workload management
Primitive controls to
throttle
Tight coupling owing
to Direct DB connect
ETL
Alarming growth in log parsing/
processing
SMP multi-threading
bottlenecks
high
medium
low
• SMP Database Application Clusters
• Non-RI DB models (RI checks at ETL layer)
• Robust hardware
• Scale quick in a non-intrusive manner
• Multi-threaded process
• Primitive workload management
• Horizontal scalability
• Mostly direct DB connect data capture
• Extensive log parsing / processing
• Heavy lifting at ETL layer
TECH OPTION 2
Managing load window/
customer expectations
TECH OPTION 2 - CHALLENGES
Capacity expansion
does not yield linear
growth
SMP multi-threading
bottlenecks
Primitive mixed workload
management
Primitive controls to
throttle traffic
Tight coupling due to
Direct DB connect
ETL
Alarming growth in log
parsing/processing
high
medium
low
TECH OPTION 3
Dual Interconnects
• MPP Database engine
• Massively Parallel Processing
• Bulk Load capability
• Sophisticated mixed workload mgmt
• Advanced throttling capabilities
• Horizontal scalability with linear gain
• Push down optimization at DI layer
• Log processing moved to Hadoop
• Heavy lifting at ETL layer
up to 1,024
nodes
Logical units
of storage
Multiple
virtual units
of work
TECH OPTION 3 - CHALLENGES
Economies of scale,
hard to justify ROI
Vendor lock down for
growth & maintenance
Large backups,
slow restores
Tight coupling due to
Direct DB connect ETL
Data share between
Hadoop and structured
environment
Managing load
window / customer
expectations
Data Appliances
mitigate to some
extent
high
medium
low
Capture
Layer
Servers
File / DB
Ingestion
Unstructured
Layer
Hadoop Tech
Stack and
Storage
Relational
MPP Layer
Data
Warehouse
UX Layer
BI & Reporting
Near-Real Time / In-Memory
Action /
Feedback /
Loopback
High-Fidelity Messaging Layer
TECH OPTION 4
TECH OPTION 4 - CHALLENGES
Switch from traditional
IT operating model to
engineering operating
model
Restructuring / retraining
team to enable use the
techstack
Agile delivery now crosses
boundaries of technology
Lack of sophisticated share
between Cloud and internal
properties
Lack of workload
management across
structured & unstructured
tools / technologies
Managing customer
expectations (prioritize
RT vs. NRT)
Growing team of
Engineers and
Data Scientists
high
medium
low
• Business problem you are trying to solve with measurable ROI?
• Is the organization ready for Agile + Big Data?
• To some extent, both are culture change (Build vs. Buy)
• Clear thinking on technology bridges
• Structured/Semi-Structured/Unstructured – Share across
• Workload Management / Traceability end to end
• Throttling across the board
• Sustainability and Scalability
• Technology - Seldom is a solution but a means to achieve it
• For success, never base your decision on Hype/Coolness/Cost
BROAD THINKING … BEFORE BIG DATA
Q&A Janet Cinfio
Alex Ignatius
http://jobs.ea.com
https://twitter.com/EA
http://store.origin.com