Scheduling control apparatus, scheduling control method and program for minimizing total cost in service system
View Patent ↗A scheduling control device performs scheduling that includes generating a directed graph having a source node, n layers of nodes, each layer of nodes of the n layers of nodes consisting of k nodes respectively corresponding to k job processing devices, and a destination node, a cost required for processing a job being allocated to each node of the n layers of nodes as a weight, a cost required for migrating the job from one of job processing devices to another one of the job processing devices being allocated as a weight on an edge; and calculating a shortest path from the source node to the destination node in the directed graph as a result of scheduling.
1. A scheduling control device that performs scheduling for locating a job in a job processing device so as to minimize a total cost in a service system in which a predetermined period is divided into n time slots based on timing at which a cost required for job processing is changed, the service system including k job processing devices, the scheduling control device comprising:
a processor; and
a memory that includes instructions, which when executed, cause the processor to execute the following steps:
collecting cost plan information of each job processing device, said cost plan information including parameter information that transitions according to the time, said parameter information including a cost required by each job processing device for the job processing for each of the n time slots,
generating a directed graph having a source node, n layers of nodes, each layer of nodes of the n layers of nodes consisting of k nodes respectively corresponding to the k job processing devices, and a destination node, a cost required for processing the job being allocated to each node of the n layers of nodes as a weight, a cost required for migrating the job from one of the job processing devices to another one of the job processing devices being allocated as a weight on an edge;
calculating a shortest path from the source node to the destination node in the directed graph as a result of scheduling, and
transmitting an instruction based on the shortest path to an orchestrator that manages a virtual network environment in which the cost in each job processing device varies according to the time slot and control the orchestrator to execute live migration of the each job processing device at the timing at which the cost required for job processing is changed based on instruction.
2. The scheduling control device according to claim 1 , wherein each of the k job processing devices is a data center, the cost required for job processing is an electricity charge, the total cost is a total electricity charge, the locating the job in the job processing device is locating a server in a data center, the processing the job is operating the server, and the migrating the job from the one of the job processing devices to the another one of the job processing devices is live migration of the server from one of the data centers to another one of the data centers.
3. The scheduling control device according to claim 2 , comprising: wherein the steps executed by the processor further includes estimating power consumption to be used for calculating the electricity charge to be allocated to the node or the edge by calculating an average of power consumption while the server is operating in a past certain fixed period.
4. The scheduling control device according to claim 2 , wherein the steps executed by the processor further includes, upon detecting that a facility amount is insufficient for a facility amount required for providing a service in the service system, performing scheduling for the facility amount that the service system has, and, for the insufficient facility amount, generating a directed graph in which a node or an edge to be a bottleneck has been removed, and calculating the shortest path from the source node to the destination node in the directed graph as a result of scheduling for the insufficient facility amount.
5. A scheduling control method executed by a scheduling control device that performs scheduling for locating a job in a job processing device so as to minimize a total cost in a service system in which a predetermined period is divided into n time slots based on timing at which a cost required for job processing is changed, the service system including k job processing devices, the scheduling control method comprising:
collecting cost plan information of each job processing device, said cost plan information including parameter information that transitions according to the time, said parameter information including a cost required by each job processing device for the job processing for each of the n time slots,
generating a directed graph having a source node, n layers of nodes, each layer of nodes of the n layers of nodes consisting of k nodes respectively corresponding to the k job processing devices, and a destination node, a cost required for processing the job being allocated to each node of the n layers of nodes as a weight, a cost required for migrating the job from one of the job processing devices to another one of the job processing devices being allocated as a weight on an edge;
calculating a shortest path from the source node to the destination node in the directed graph as a result of scheduling, and
transmitting an instruction based on the shortest path to an orchestrator that manages a virtual network environment in which the cost in each job processing device varies according to the time slot and control the orchestrator to execute live migration of the each job processing device at the timing at which the cost required for job processing is changed based on instruction.
6. A non-transitory storage medium storing a program which, when executed by a scheduling control device that performs scheduling for locating a job in a job processing device so as to minimize a total cost in a service system in which a predetermined period is divided into n time slots based on timing at which a cost required for job processing is changed, the service system including k job processing devices, causes the scheduling control device to execute a scheduling control method comprising:
collecting cost plan information of each job processing device, said cost plan information including parameter information that transitions according to the time, said parameter information including a cost required by each job processing device for the job processing for each of the n time slots,
generating a directed graph having a source node, n layers of nodes, each layer of nodes of the n layers of nodes consisting of k nodes respectively corresponding to the k job processing devices, and a destination node, a cost required for processing the job being allocated to each node of the n layers of nodes as a weight, a cost required for migrating the job from one of the job processing devices to another one of the job processing devices being allocated as a weight on an edge;
calculating a shortest path from the source node to the destination node in the directed graph as a result of scheduling, and
transmitting an instruction based on the shortest path to an orchestrator that manages a virtual network environment in which the cost in each job processing device varies according to the time slot and control the orchestrator to execute live migration of the each job processing device at the timing at which the cost required for job processing is changed based on instruction.