1. Read path - Client, Router, Storage Node
a. Latency - streaming, GC, code-gen, memory & connection management
b. Stability - capacity planning, high availability, back pressure, etc.
c. Deep collaboration with People You May Know (and other AI teams) - AI integration, read computation, online deep learning.
2. Write Path - Hadoop-to-Venice, Samza-to-Venice, ETL, Storage Node
a. Da Vinci - venture bet to bring Venice ingestion capabilities to other systems.
b. Write Path features - active-active replication, partial updates, storage quota.
c. Deep collaboration with Waterloo (standardization infra) - ETL to Hadoop, reprocessing, usability improvements.
3. Control Path - Controller, Admin tool, Storage Node
a. Operability - improve the Venice Health Dashboard, eliminate single points of failure, kill superfluous components, enhance fault detection and self-healing capabilities
b. Deep collaboration with Nuage - self-service onboarding for more features
c. Deep collaboration with Helix - Helix-as-a-Service integration, rack-awareness, design review and adoption of many new Helix features
a. Performance - profiling and tuning
b. Cloud native - elasticity, enable wider range of performance-cost tradeoffs, etc.
