A ThinClient Approach to RealTime Market Data

1
A Thin-Client Approach to Real-Time Market Data

• Presentation at
• Wall Street on JavaTM Technology
• February 5, 2003

Steve Shum Executive Vice President Sector Data,
LLC
2
Objective

• Build a powerful yet cost-effective application
  for real-time streaming quotes and analytics

3
Requirements

• Advanced, institutional-class feature set
• Streaming watch lists, NASDAQ Level II
  (SuperMontage), Time Sales, charts, news,
  tickers, option analytics
• Multiple real-time components updating
  simultaneously
• Highly flexible and highly customizable user
  interface
• Multi-monitor support
• Professional-level performance
• Small (
• Complete mobility of user configuration and data
• Seamless software upgrades

4
Why Java?

• Swing library pre-built components enable much
  faster development
• Multi-Threading powerful feature of Java
• Garbage Collection
• Cross-platform support single version of
  application runs on Windows, UNIX, Linux, and Mac

5
SectorEngineTM

• 100 pure Java application built on Swing library
  (J2EE J2SE)
• Operates within standard browser window OR as
  stand-alone Java application using Java Web Start
  
• Features, functions, and performance comparable
  to fat-client professional market data products
• In commercial use at over 80 customers, including
  floor traders at Pacific Stock Exchange

6
Technical Overview
7
Client Side

• Thin-client model
• All user configuration settings and data are
  saved on server
• Client serves as GUI interface with essential
  application logic only
• Core logic runs as services on server
• Client accessible from anywhere with Internet
  connection
• Requires J2SE plug-in v1.4.0 or higher

8
Application Windows

• Client workspace utilizes internal (floating)
  windows to provide data quotes, news, etc.
• Multiple instances of same window type supported
• No limit on number of windows except client
  hardware capacity
• Windows management within workspace
• Configuration Manager User customizations
  (window size position, tickers, fonts colors,
  etc.)
• Communication Manager Data exchange and event
  notification between windows
• User settings and communication rules described
  in XML and saved on server

9
Server Connection

• Static data HTTP tunneling
• Dynamic data Single HTTP connection on port 80
• Dynamic subscription ID passed from server to
  client through serialization
• Connection kept alive by periodic messages
  (pulses) sent by server
• Port issues resolved by implementing Apache
  module on server

10
Data Handling

• No data stored locally on client side at any time
• Data flushed and reloaded when windows are opened
  or closed
• For dynamic windows, requested data cached and
  updated dynamically
• Most static data computed and formatted on server
  side and displayed in client upon request

11
Streaming Real-Time Data

• Dynamic windows request feed data types (quotes,
  trades, news, etc.) from server
• Subscription Server
• Receives all data subscription requests from
  client
• Registers unique subscriptions only
• Pushes updates for registered data items to
  client
• Minimizes bandwidth
• Feed Manager
• Distributes data updates to multiple client
  windows
• Tracks which windows are registered for which
  feed data types and identifiers (ticker symbols)

12
Performance - Issues

• Problem As a higher-level language, Java tends
  to consume more memory and CPU resources
  particularly in dynamic applications where
  significant screen updating (repainting) is
  required
• Challenge Make SectorEngine perform as well as
  traditional fat-client applications

13
Performance - Solutions

• Sector Data
• Feed Manager critical for
• Handling real-time streaming data distribution
• Optimizing performance of client
• Performance gain approximately 20
• Sun
• Improvements made in version 1.4 of plug-in
• Better garbage collection
• Better virtual machine
• Lighter Swing library J Table, Paint Manager
• Performance gain approximately 15

14
Plug-in Issues - Availability

• Vast majority of potential users do not have Java
  plug-in installed
• Larger financial institutions firewalls block
  download
• IT staff involvement is usually required to
  install

15
Plug-in Issues - Updates

• Process for updating plug-in is cumbersome
• Requires manual un-install of previous version
• Will be addressed in v1.4.2
• Inter-operability problems arise between
  versions
• Customers may have applications running on
  previous versions that will not run on v1.4
• Known issue

16
QA

• Thank you!
• Steve Shum
• Executive Vice President
• Sector Data, LLC
• sshum_at_sectordata.com
• 415-782-0454 x151

Sector Data, LLC, is not affiliated with Sector,
Inc., or SIAC