Information Technology Reference
In-Depth Information
The complex job descriptions are generated
with the multiple descriptions with mutation al-
gorithm. The algorithm differs from the multiple
descriptions approach, in that it considers, that in
such cases if the sections of the new and the “to-
be-replaced” description are the same and only
their attributes differ, they presumably reflect the
execution of the same sequence of operations.
The different attributes indicate that the actual
parameters were slightly different, however it
cannot be said that either the new or the old
description is closer to reality. Therefore, the to-
be-replaced description should not be replaced but
only mutated : the attributes of the sections have
to be recalculated using the new attribute values
(e.g. their average can be used) as determined by
the algorithm presented by the single generalized
description approach. The mutation operation
is defined by the single generalized description
approach.
After the execution of a given job the complex
job description is updated as follows:
scription, then the old composed description
is used , and the absolute weight of the ”clos-
est” member job description is increased by
1.
If the number of member job descriptions is
limited , then mutation can be used to keep the
number of member descriptions under the limit,
and also to preserve the knowledge carried by the
new individual job description. The algorithm
resulting in complex job description with limited
siye is as follows,
1. If the number of member job descriptions is
less than the limit, the previously presented
algorithm is used.
2. If the number of member job descriptions
already reached the limit, this approach will
mutate (using the algorithm presented in sec-
tion Generating complex job descriptions)
the member job description which is the
closest to the new individual job description
according to the similarity measure. The
absolute weight of the mutated member job
description is increased by 1.
1. If the distance between the newly generated
individual job description and the member
job descriptions is greater than a threshold,
the new individual description is inserted
into the composed description. The absolute
weight of the new member description will
be 1.
2. If there is a member job description which
is closer to the new individual description
than the threshold value, and the diversity
of the member descriptions would increase
with the insertion of the new description,
then the closest member job description is
mutated using the algorithm presented in
section Generating complex job descriptions.
The absolute weight of the mutated member
description is increased by 1.
3. If at least one of the complex description
members is closer to the new job description
than the threshold, and the diversity of the
composed description would not increase
with the insertion of the new individual de-
SCHEDULING STRATEGIES
This section will present the proposed schedul-
ing strategies that exploit the information stored
by the job descriptions. The major difference
between the scheduling strategies is that while
the first, static data feeder , strategy prepares the
input files before the job would be executed, the
second, dynamic data feeder , strategy delivers
the necessary files parallel to the execution of
the job, in a just-in-time manner. The process of
job scheduling and execution comprises of the
following major steps.
1. The user submits the job and its description.
2. The system looks up the corresponding job
behaviour description using the description
repository service .
Search WWH ::




Custom Search