Information Technology Reference
In-Depth Information
The operation of the system defined in this way is the calculation of the task
consisting of
Nt
partial tasks. The resource, distribution of which was balanced in the
environment, was made up of partial tasks intended for the calculation in respective
processors. During the calculations the number of partial tasks varied since those
tasks were successively generated, which resulted in an increase in the resource. On
the other hand, when the calculations of partial tasks were realized, their number
dropped.
These tasks were generated according to the following scenario:
•
At the moment of initiating the calculation of the task a certain number of partial
tasks (referred to as initial constituting 10% of the number
Nt
in the studies) was
generated in chosen processors of the structure (constituting about 10% of all the
processors).
•
At the end of calculating each task, with certain established probability, a decision
was made as to whether to generate a certain number of partial tasks in a given
processor (node), and after completing a given task a specified number of new
tasks was generated.
•
The number of generated partial tasks was controlled in the system and after
achieving the value
Nt
the process of generating new tasks was blocked.
In the multiprocessor environment, balancing the tasks is realized through the
agent system, which consists of the following agents:
•
The agent of type
A
g
0
which exists only in one instance in each node and is
responsible for managing this node (processor).
•
The agents of type
A
g
1
search for the node (processor) which may send a certain
number of tasks.
•
The agents of type
A
g
2
search for the node (processor) that is able to receive a
certain number of tasks.
•
The agents of type
A
g
3
, also referred to as “unemployed” when they exist in the
system search for tasks to realize.
The description of the structure of the system is the same as presented in Sect.
4.2
.
As a measure of effectiveness (quality) of calculations the indicator
E
f
(
efficiency
)
was accepted in the form of:
T
c
E
f
=
T
r
,
(4.1)
n
∗
where:
T
c
is the time of calculating all partial tasks on one computer;
T
r
is the time
of performing all tasks onto the multiprocessor structure;
n
—a number of nodes
(processors) in the structure.
The indicator
E
f
is used for defining the efficiency calculation of tasks onto the
structure and refers to the calculation of a given task consisting of the specified
number of partial tasks (
Nt
).
