The Scientific World Journal / 2014 / Article / Psdc 1

Research Article

Efficient Dynamic Replication Algorithm Using Agent for Data Grid

Pseudocode 1

Pseudocode for region optimizer and subregion Optimizer.
Optimizer (region)
Inputs: Grid Topology, Bandwidth, Storage Space and Load Gauge
Outputs: Find Best Candidate for replication, Load Balancing, Job
Execution Time, Storage Usage, Network Usage, Network Utilization,
Creation of New Replicas
Method:
 (i) Job is submitted to master node of grid by the agents/user.
 (ii)Master node has the global view of all regions in the grid.
 (iii)   Initially do
 {send request to head nodes asking for
 (available bandwidth, load gauge)
 }
end;
 (iv)Find maximum available bandwidth and minimum load gauge.
 (v)  Schedule job to region.
 (vi)    Call optimizer for sub region to get best replica.
 (vii)  Execute all the jobs.
Optimizer (sub region)
Method:
 (i)Agent at head node of each region keeps track of
   stored files, availability status, and computing
   capacity of node and access frequency with in region.
 (ii) if (file exist) { process job;
   terminate optimizer;}
 (iii)else {if (request file is not available in local sub region)
   fetch from nearby sub region;
 (iv)proceed to execute the job with replica}
 (v) if {(free space available in SE of the node where
   request was initially scheduled, store new replica)
   store data;}
   else {
   (check free space in SE of nearby node within in
   sub region and replicate)
      break; }
 (vi) if (!enough free space with the sub region) {
   (sort file in least frequently accessed order)
   for (each file in sorted list) (
   if { (access frequency of new replica access
   frequency of the old file)
   delete old file;}
   if {enough free space
   store new replica;}
   else go to 6;
   break;