K. L. Shi
source s, therefore, improves the economic benefit of power grid. However, the existing grid computing task de-
composition algorithm easily run into the problem of local optimal solution, which affects the normal manage-
ment of resource request from users shows a performance of power grid. The task scheduling under the cloud
computing is collecting all kinds of information from power grid and giving a feedback to the power cloud. The
computers will calculate the relevant parameters and indicators based on the information stored in cloud re-
source s, and give valid instructions by feasible judgment of operation state both in safety and economy to
achieve a relatively stable state by necessary adjustments.
Inappropriate task scheduling strategy can lead to waste of cloud resources and even influence the quality of
power grid, which may seriously affected the enterprise’s economic benefit.
3. The Basic Principle of IHA
The basic principle is to partition computing tasks. Namely, sub-tasks are stored in one node of the and/or tree,
and the user request task decomposition means choosing the right point of division and building structure of the
tree. We estimate overall consumption after using heuristic method to find a division of the task, when dividing
point selection is improper, sub-task will not be able to perform and be regard as infinity of consumption. Re-
peat these steps until all the sub-tasks couldn’t be further subdivided, then we think the task decomposition is
completed.
Task decomposition has two important steps: first, in order to ensure the optimal solution is for overall tasks,
every step must meet the constraint conditions of relevant task decomposition; Second, in order to improve the
efficiency of resource requesting and distributing and minimize the costs, we should carry on reasonable classi-
fication and allocation to these divided tasks.
4. The Combination of Bionics
Considering some related concept of genetic engineering in biological science, gene cutting, we need use a so-
called “scalpel” to do molecule specificity cutting for some specific nucleotide sequences on the DNA, whose
scientific name is restriction endonuclease. We regard the initial power grid computing tasks as a DNA analogy,
on which there are one or some nodes of operation, monitoring, protection, distribution, marketing, etc. The
constraint to divide the task can be regard as “scalpel” to do specific cutting for the computing task which con-
forms to the corresponding node cutting constr aint s. Then we take one of these nodes to do further division with
the IHA and look for task dividing point of a node. If the node has at least one task segmentation point and the
sub -tasks can be retrieved in the cloud after segmentation, then this point can be determined as the division point
of tasks. Do further division until all sub-task can’t be divided any more, task decomposition process is over. If
the divided sub-task can’t be retrieved in the cloud, rework following process above until all its tasks can be re-
trieved in the cloud. Finally, polish sub-tasks tree according to the consumption to perfect the decomposition
process.
Based on the natural selection in biological science, we expand and perfect the original algorithm to make it
has a good performance both in reasonable decomposition but also for optimal resource allocation. We define
these alternative task scheduling schemes as a species. First of all, we encode the initialized population, then we
define a fitness function to evaluate whether the individual of population is competitive enough. The task sched-
uling scheme with highest fitness function value above all alternative schemes will give the most appropriate
scheduling command.
5. Flow Chart of the Alg ori th m
Certain grid tasks and resources must correspond to different solutions. Because the degree of emergency and
importance is directly related to the real-time requirements of tasks, thus affecting the priority level of a task in
the process of solving [4]. We first make a division of emergency degree for the sub-tasks which can’t be di-
vided any more, then a resource allocation scheme is presented. Next, to avoid the full arrangement for each task
as well as make a selection from the alternative schemes created by the full arrangement, therefore saving a lot
of time and computing resources to improve the efficiency and saving resources. Finally, we take further
screening of these relatively reasonable scheduling scheme by using the theory of natural selection and define
the fitness function considering time and cost to give a standard to screen out optimal solution as well as give
scheduling command [5]. The process is shown in Fig ure 1.
