Online Prediction of the Running Time of Tasks
Cluster Computing
A computational economy for grid computing and its implementation in the Nimrod-G resource broker
Future Generation Computer Systems - Grid computing: Towards a new computing infrastructure
Design and Evaluation of a Resource Selection Framework for Grid Applications
HPDC '02 Proceedings of the 11th IEEE International Symposium on High Performance Distributed Computing
| Hi-index | 0.00 |
In order to charge the computer power in the grid, we have made an effort to work towards a standard pricing unit. Currently there is a lot of work done towards the development of grid economy models and architectures. But when it comes to charging, the usual metric which has been popularly used is $ per CPU per Hour which seems to be too simple. Our effort is to make this metric more meaningful to both grid service provider and client. We argue that in a particular grid host the metric should reflect the true load consumed by the clients and the delays caused due to the other loads. Further it should eventually reflect the network, memory etc consumed by the client as well. Previously we have studied about the prediction in the grid after introducing the division of the load average at the kernel level. This gave more meaning to the historical load collection as CPU historical load data had been collected separately for each login user. Interestingly, later on the division of load strategy has been helpful in the development of a meaningful tariff mechanism and would be demonstrated in this paper. Eventually this fare mechanism would be used to predict the computational costs, which would certainly contribute to the scheduling in the grid.