1 the netflix data platform – a recipe for high business impact kurt brown, director, data...
TRANSCRIPT
1
The Netflix Data Platform – A Recipe for High Business ImpactKurt Brown, Director, Data Platform
Agenda
• Architecture
• Cloud
• OSS
• Philosophy
Architecture
S3
~5 PB
S3
vs.
Sting
S3
Sting
Data Platform
Tez
/ Shark
?
S3
Data Platform
TezYarn
???
Sting
Cloud
Yes
OSS
Cost
Cost
Flexibility
• Growth
• Spikes
• Self-service
Services
• SQS, EMR, S3,…
• Available
• Focus!
Redundancy / Global Presence
• Multi-DC
• International
Cloud Native
• Build for failure
• Swing systems
• Excess capacity
Right for everyone?
• Are you Google or Facebook?
• Netflix
• Undifferentiated
• Opportunity Cost
Cost?
OSS
General OSS
• Pooled goodness
• Manage own destiny
• Vendor relationships
• Conflicting agendas
Netflix OSS
• Meet Netflix needs
• Shared direction
• Recruiting & Retention
• Code quality
• Contributions
Data Platform OSS
• Genie
• Lipstick
• Aegisthus
• S3mper
• Suro
• Future: Inviso
Should you open source?
• We’re all using it
• You get value
• Others would get value
• Will invest some time
Philosophy
Netflix Expense Policy
“Act in Netflix’s Best Interest”
Development & Deployment Flow
BestPractices Components
ConsultationTools
Automation
Clean-up
DBAs?
QA Team?
Testing
Accept things WILL break
Safety Nets
Give folks what they need
Eliminate Process
• Status Reports
• Tickets
• Documentation
Tell vs. Ask
Does it really matter?
Questions?
http://jobs.netflix.com [email protected]