Browse Prior Art Database

Process Control Executive - A New Paradigm for Operating System Design

IP.com Disclosure Number: IPCOM000100732D
Original Publication Date: 1990-May-01
Included in the Prior Art Database: 2005-Mar-16
Document File: 6 page(s) / 281K

Publishing Venue

IBM

Related People

Bozman, GP: AUTHOR [+4]

Abstract

Disclosed is a new process control executive (PCE). It is a new paradigm for constructing operating systems and/or system control programs that are required to function efficiently in highly parallel environments. It solves problems relating parallel processing in both high-end and low-end computing systems. Among these problems are: a) hardware and software resource management; b) contention and concurrency control; c) MP inefficiencies relating to queueing, dispatching, and serialization.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 23% of the total text.

Process Control Executive - A New Paradigm for Operating System Design

       Disclosed is a new process control executive (PCE). It is
a new paradigm for constructing operating systems and/or system
control programs that are required to function efficiently in highly
parallel environments. It solves problems relating parallel
processing in both high-end and low-end computing systems. Among
these problems are: a) hardware and software resource management; b)
contention and concurrency control; c) MP inefficiencies relating to
queueing, dispatching, and serialization.

      Using the PCE method, all modules that handle a specific
request pass work via a packet known as the Resource Request (RRQ).
The RRQ completely defines a unit of work in the system. It contains
and/or points to all the data necessary for its processing. These are
typically parameters, save areas and/or stack, priority, and routing
information (both local and remote). It is created by some initiating
resource manager, routed by the PCE to a server resource manager,
processed, and eventually destroyed as a result of some terminal
condition.

      The RRQ is used for all resource-manager calls. This approach
allows: a) queueing effects to be managed; b) the state of the system
to be easily interpreted; c) a scheme analogous to message passing to
solve concurrency problems (1); d) piping capabilities (based on the
route list in the RRQ); e) conditional re-routing (by changing the
route list or some of its entries).

      RRQs are queued on control structures that define resources
(i.e., system services or devices). In the PCE, a resource is defined
by a Resource Control Block (RCB). These are grouped together in
Resource Partition (data structures). They are accessed by the CPU
through addressing vectors (Resource Partition Vectors) and a
Partition State Table. The process is described in the figure.

      Each resource request is processed by one or more resource
managers. A resource manager has several fixed attributes:
 o  Serial or non-serial. A serial resource manager can be active
with only one RRQ at a time. A non-serial resource manager can
service as many RRQs concurrently as there are available processors.
 o  Requires stack storage. The stack is appended to a RRQ.

      In addition, a resource manager has the following dynamic
attributes:
o  Ready:  A resource manager that is not ready is blocked and cannot
provide service.  It might, for example, represent a device that has
been varied off-line.  The normal state of a resource manager is
ready.
o  Active:  In this state the resource manager is processing an RRQ.
o  Inactive:  This state implies that the resource manager is avail
 able to service a RRQ.
All of the dynamic state transitions are controlled by the PCE.

      In general, what may be considered a logically discrete
resource request (e.g., an I/O server) may be composed of more than
one RCB (e.g., an I/O ini...