Developing the Bloomberg Terminal - Local Performance & Measurement Techniques

Explore techniques for developing and instrumenting a high-performance, near-real-time multi-window browser system in this conference talk from NDC Conferences. Dive into the architecture of the Bloomberg Terminal, combining proprietary C++ data processing technology with an embedded Chromium browser core and V8 JavaScript engine. Learn about server-side components, application container processes, metrics subsystems, and data layer optimizations. Discover strategies for efficient memory management, including custom allocators and polymorphic allocators. Examine performance improvements through portable columns, worksheet expressions, and coalescing. Investigate UI bindings, challenges in minimizing garbage collection, and techniques for measuring responsiveness. Gain insights into JavaScript levels, Chrome Dev Tools Inspector, and UI timing. Finally, explore the integration of WebAssembly and its C toolchain within this complex distributed run-time environment.

Introduction
Agenda
Background
What is it
Server side
Backend
Lua scripting engine
Application container processes
Metrics subsystem
Data layer
Data layer components
Combining data
DataNow Inspector
Datum
Custom data types
What is datum
Can we do better
Floatingpoint format
String optimization
Polymorphic allocators
How does this work
Custom allocator
Monotonic Buffer
Data Layer Performance
Portable Columns
Worksheet Expressions
Coalescing
Chromium
Application Container
UI
Bindings
Challenges
Minimize GC
Life cycle
Application
Proxy object
Disposal pattern
Overall CPU load
Bloomberg app
Bloomberg 3D view
Memory usage
Cake view
Live memory usage
Measuring responsiveness
Instant message system
JavaScript levels
Chrome Dev Tools Inspector
UI Timing
Playable
Recording
Free integration testing
WebAssembly
What is WebAssembly
C toolchain
How do we hook up to it