Information Technology Reference
In-Depth Information
means that the client communicates with an elastic pool of workflow server instances on
the cloud side, which is different from the traditional client-server model where the
client communicates with one static or a cluster of physical workflow servers. The high
throughput, elastic scalability and cost-effectiveness have been taken account into the
system, so it can achieve a dynamic, elastic, and sustainable scalability.
5
Problem Analysis and Prospect
5.1
Existing Problems
The researchers have made great progress in Workflow Scheduling Algorithms so far.
But there are also some problems in them, as follows:
1. Most workflow scheduling algorithms focus on the cost and the deadline of
workflow, but other scheduling parameters have obtained little attention, for exam-
ple, like resource utilization, resource reliability, fault-tolerant and so on. The re-
source in cloud is huge, not only we should pay attention to meeting the deadline
but also the improvement of singe resource utilization. At this time, we should
solve the problem when the execution happens faults, too.
2. Lack of the workflow task model in cloud. Always, it is more practicability when
workflow scheduling algorithms have been made verification. But in the current
cloud environment, most researchers just use simple mathematical model or di-
rected acyclic graph because of lacking task log. Most task models are not made
according to the specific operational tasks in cloud, the experimental results are
least persuasion, so the problem of making appropriate model in cloud needs to be
resolved urgently.
3. Most workflow scheduling algorithms have been implemented in cloud simulators,
besides the scientific workflow. We must apply workflow scheduling algorithms in
real-life environment so that can solve more real practical problem.
5.2
Prospect
1. The reliability should been considered in workflow scheduling algorithms, the ro-
bustness can have a significant impact on experiment. Taking the reliability and
robustness into account can close to real situation.
2. Duplication based on workflow scheduling algorithm. The parts of tasks can be au-
tomatically duplicated to the free computing nodes which have some savings.
3. To be combined with the practical application. Today, most workflow scheduling
are used in simulation environment, while it still has some deficiencies compared
with real environment.
6
Conclusion
The optimization of cloud workflow scheduling can make full use of various cloud
services, which will greatly promote the execution of a workflow. Workflow
scheduling is the most important problem in the execution of workflow management
Search WWH ::




Custom Search