11 September 2003

1
Real-Time Database Brief

• Mars/03-09-09

2
Some of the RT DB folks

• Mayford Roark Inventor of CODASYL-based Real
  Time Data Manager (RTDM) CSCI for Seawolf
  submarine combat system, author of SRS for
  Functionally Integrated Resource Manager (FIRM)
  real-time object-oriented database manager.
  Represented Lockheed Martin on US Navy NGCR DISWG
  standards body, published authoritative paper on
  object-oriented real-time data management for
  1996 Software Technology Conference.
• Steve Ruberg Former employee of Air Force
  Research Laboratory (AFRL) Wright Laboratory who
  envisioned FIRM and guided the FIRM RFP and its
  AFRL requirements
• Mike Card (me) Member of Mayfords RTDM
  development team, software lead for FIRM program.
  Represented Lockheed Martin as voting member of
  the now defunct Object Data Management Group
  (ODMG) and as member of ECMA Technical Committee
  40 for Ada95 object database bindings to ODMG
  object model (with extensions to support
  real-time DB). Software lead for Lockheed
  Martins RODEO (Real-Time Object Database with
  Extensions to ODMG), successor to FIRM program.
• Paul Springer Real-time database application
  developer for Seawolf program, member of FIRM and
  RODEO software development teams. Developed
  multithreaded socket server for FIRM/RODEO and
  designed RODEO replication mechanism. Application
  software development lead for Advanced Integrated
  Electronic Warfare System (AIEWS), which used
  RODEO as track file manager and for emitter
  classification, both very demanding firm
  real-time applications

3
Real-time DB timeline
RTDM
EagleSpeed-RTDB
FIRM
RODEO
OMG?
ODMG-Ada95
NGCR DISWG
8/2001
9/2003
8/1991
4/1/1993
9/1995
8/1998
11/1996
11/1998
NGCR DISWG US Navys Next Generation Computer
Resources Database
Interface Standards Working Group
4
Why did this stuff ever start?

• 1987 - AN/BSY-2 Combat Sonar Systems
  Requirement
• The Database Manager shall provide relational or
  equivalent access as well as
• direct access to underlying tables
  (3.2.1.2.3.2.1)2.
• Selected COTS DBMS product and ran benchmarks
  only to find that
• Performance didnt meet real-time
  requirements.
• Execution times were unpredictable.
• DB Administration was required every 24 hours
  for storage defragmentations and garbage
  collection
• DBMS did not guarantee contiguous storage on
  disk, resulting in slower data access.
• Type conversion from Ada to DBMS and vice
  versa resulted in additional complexity and
  development costs.
• After exhaustive attempts to utilize a COTS DBMS,
  the decision was made to
• design, build and field our own Real-Time DBMS
  for the AN/BSY-2 Combat System.

5
Conventional vs. real-time DB
6
Essential RT DB Features

• Largely fragmentation free, supporting
  long-term unattended
• operation.
• Minimizes Interprocess Communication.
• Uses appropriate Data Model.
• Utilizes unique data structures.
• Uses Main Memory.
• Contains efficient Concurrency Control
  mechanism.
• Provides fault-tolerance.
• Worst-case performance times can be predicted
  at design time
• Real-time characteristics must designed into the
  architecture of
• the DBMS they can not be added onto an existing
  DBMS.

7
Data Model

• The network (CODASYL, Conference On DAta SYstems
  and Languages)
• and object (ODMG) data models were excellent for
  use in real-time
• embedded systems because they allowed the
  application developer to
• control the navigation path for accessing the
  data, thus making determinism
• possible.
• In addition to providing control of the data
  navigation path, the object data
• model also allowed more seamless integration of
  the programming
• languages type system with the database engine.
  This was a double
• boost for real-time application developer
  productivity.
• It was our recognition of these benefits and
  AFRLs desire to support
• standards that led to the FIRM/RODEO teams
  participation in ODMG.

8
Real-time DB Benefits

• Unlike traditional home-brewed approaches to
  this problem based on
• user-defined data structures, mutexes and
  messages, a real-time database
• engine offers a common data management solution
  that provides fast
• data structures and key lookups, transaction
  semantics, fault-tolerance,
• persistence and replication.
• A custom-brewed solution might run faster, but
  a real-time database
• solution will be faster to build and get
  working. It will also make it easier
• for programmers to work in different areas of a
  large, complex project. It
• will allow the application developers to focus on
  the application, not on
• managing the data.

9
What was actually developed?

• Do excerpts from RTAG brief(s) and
• STC presentations if there is time /
• if desired