Information Technology Reference
In-Depth Information
failed for any reason, thus, MA will update the database file (there is a Boolean field in
table that if it is set to TRUE, this means that the related job has been finished success-
fully, otherwise, it means that the job is not successfully done and has failed). Also, we
have considered some new tasks for Grid Resource Broker (GRB), which we have
called Optimal GRB. Before selecting any nodes (for aimed job) by GRB, one of these
tasks will be executed, this is responsible for sending a packet to each node on grid
besides previous tasks. Needless to say, this task can be executed during recourses dis-
covery operation by GRB. Further, as already stated, there are many different methods
to find resources (nodes), but will not concentrate on how we can discovery nodes; and
we will not mention them in this paper. Suppose that, there are many different nodes in
our grid that are ready for executing jobs and we want to select some nodes in the pool
of these nodes. At the beginning, GRB has sent a packet to each connected nodes to our
grid. This packet contains some information about a new job (e.g. IP Sender, Size of the
job, Size of needed RAM and HDD, Estimation time needed for execution, approximate
execution start time, minimum power needs to CPU, OS status, etc.).
On the other side, when MA in node gets this packet, it will open the packet for
analysis. If there are sufficient conditions (internal resources node) to do the desired
job, MA will perform a data processing technique on its own mini-database (or its log
file) to obtain some computation for this job. In many conditions if broker send a Data
mining request to MA, the MA application will execute a DT algorithm and then it
will send useful rules to broker. Some of produced results are as follows:
•
Average Hit Ratio (AHR): This attribute represents an average rate of success in
all previous times.
•
Number of all submitted jobs on this node (AAJ).
•
Number of all jobs submitted at this time, on the previous days, on this node
(AATPJ).
•
Number of all jobs successfully finished at this time, on the previous days (NSTP).
•
Hit Ratio for this time-period on previous days (HRTP). For example, how many
jobs in 1.30 AM o'clock to 2.00 o'clock have been executed?
•
Average Size of successfully finished jobs (ASF).
Average Response Time for finished jobs (ART).
•
•
Average Response Time for jobs that have the same size as the purposed job and
have been successfully finished (ARTSS).
Hit Ratio for the last twenty jobs (HRT).
•
•
Date, Time and Size of the last successfully finished job (LSJ).
•
Date, Time and Size of the last failed job (LFJ).
•
Size of the largest successfully finished job (LSI).
•
Numbers of all previous jobs that almost have the same size as the purposed job
(ASS). Needless to say, the size of the previous jobs is not exactly the same as the
size of the desired jobs. For example, for a job with size=340KB we must find all
of the previous jobs between 1K to 500KB size.
•
Number of all previous jobs that have the same size as the purposed job and are
successfully finished (NSS).
•
Moreover, processor speed and CPU availability (Idleness) are important for
choosing a node.
Search WWH ::
Custom Search