A Replica Based Co-Scheduler (RBS) for Fault Tolerant Computational Grid - Cloud, Grid and High Performance Computing: Emerging Applications - page 115

Figure 2. Reliability v/s Number of Sick Nodes Failing

Figure 3. Turnaround Time With and Without Node Replication

been introduced resulting in better fault tolerance
on moderate cost. So a better cost-performance
ratio is achieved.
The job of the RBS begins when the job of the
main scheduler, responsible for selection of the
cluster for job execution, finishes. Performance
of RBS is analyzed by considering its integration
with the TSM, which is a GA based scheduler

proposing an allocation for the job modules which
results in the minimum turnaround time offered to
the job. For the cluster, selected by the TSM, the
RBS evaluates the vulnerability of the nodes on
which an allocation has been made by comparing
their failure rates λ lt with some threshold failure
rate λ th . Selection of the thresholds depends on the
domain knowledge of the cluster along with the