Light-Weight Load Management Protocol based on Reservation and Feedback Loops

Anna Schwanengel and Gerald Kaefer


Load Balancing, Cloud Computing, Protocol Design, Resource Allocation


High performance computing and clustered server farms offer solutions for time-consuming and storage-intensive computing jobs. Albeit cloud computing comes up with theoretical infinite resource capacity and enables elasticity on computing basis, also new challenges emerge. Efficient dynamic resource management based on automated physical instance allocation is key in cloud environments. Though appropriate techniques are offered by vendors, it is only possible to react to load through manual configuration and no automatism exists. Besides, booting new instances means administrative and organizational efforts and supplementary costs, because the instance allocation delay is not infinitely small, but rather about several minutes today. Scaling required resources up and down should be well-planned in order to avoid unnecessary provisioning costs. We present a mechanism to effectively manage varying loads of services based on a light-weight protocol, appropriate adapters and a management component with regard to SLAs, costs, and provisioning time.

Important Links:

Go Back