This document describes a multiprocessor database architecture for real-time systems. The architecture uses a main memory database to meet strict timing constraints of tens of microseconds for control tasks. It supports direct, concurrent shared memory access and allows developers to specify real-time constraints like access times within application code. Remote database access is also supported through remote procedure calls while hiding latency.

2.  Hard real time system
 Strict timing constraints
 High performance
 Guaranteed response time constraints
 So, the entire database sits in the main memory

3.  Multiprocessor Database Architecture for Real
Time System
 This is for hard real time systems
 Main memory database
 Object oriented (C++ elements)
 Support direct, concurrent, shared memory data access
 Support explicit declaration of real time constraints and
semantic constraints within the application code
 Constraint specification like Access time
 Write < 80 micro sec
 Read < 50 micro sec
 Persistance -> volatile

5.  Suitable for NGC (Next generation machine
controllers) for automated factories
 Transaction timing in the order of tens of
microseconds
 Control task
 Periodic task
 Extracts current sensor values from the database and
computes new control signal for the actuators

6.  Data accessed extremely high speed must be stored in the
physical shared memory as the virtual memory and disk
based databases may generate the page faults
 Applications need not know whether the database access
is local or remote. The database handler hides the
information of being remote. Remote access achieved
through Remote Procedure calls (RPC).
 There may be some communication delay when the data
is accessed remotely. The shared data manager (SDM)
tracks the location and identity of the shared memory
objects and also constructs its own database handle for
each object to service remote requests.