Systems Network Architecture Distribution Services Agent In-Progress Queue Methods and Recovery
Original Publication Date: 1995-Feb-01
Included in the Prior Art Database: 2005-Mar-30
Forman, RF: AUTHOR [+3]
Disclosed are several related techniques for high-integrity queueing, recovery, tracking, and optimization.
Systems Network Architecture Distribution Services Agent
Queue Methods and Recovery
several related techniques for high-integrity
queueing, recovery, tracking, and optimization.
Systems Network Architecture Distribution Services
(SNADS) format set 2 uses high-integrity queueing techniques
its implementation. Two representative examples of these techniques
its implementation and handling of the agent in-progress queues and
the concept of locked queue entries.
In-Progress Queues - SNADS support provides a way for
the agent to send and receive distributions in a sequence of steps.
sequence of steps provides higher integrity for distributions than
single send or receive operations would provide.
that arises is one of memory loss when the agent is
sending a distribution. Because of the high integrity exchange
between the agent and SNADS to distribute a distribution, the agent
must issue a sequence of operations to send a distribution rather
than just one send command. The agent could possibly lose its memory
between operations. This loss of memory could mean that the agent
has forgotten what operation it just issued for a particular
distribution, or the agent could have completely forgotten about the
distribution itself. Since SNADS is the provider of the send/receive
interface, it was deemed necessary for SNADS to track distributions
being sent in order to prevent memory loss.
Note: The SNADS
receive interface does not incur the problem of
send interface uses an internal logical queue to keep
track of distributions being sent by the agent. This queue is
referred to as the "agent in-progress" queue.
Distribution - The tracking of a distribution is
done via distribution states which tell SNADS where the agent is in
the sequence of operations when the agent is sending a distribution.
The following is a list of the distribution states:
4. Ready To Be Routed
how this tracking works, it is necessary to
understand the job structure of the agent and SNADS functions. When
the agent invokes a SNADS function to perform a send sequence
operation, the SNADS function runs in the same job as the agent. For
each send sequence operation, the SNADS function will change the
state of a distribution at least once. The state of the distribution
is stored in the distribution itself which resides on the agent
in-progress queue. Since distributions on the agent in-progress
queue are permanent spaces, they will still exist even if the agent
job ends in the middle of a sending sequence. Consequently, the
state of a distribution is never lost.
If an agent
job starts up again after an abnormal failure and
invokes another SNADS send sequence function, SNADS will determine