Distribution and load balancing strategies in private cloud monitoring

The growth of private clouds causes that more efforts be needed to manage this type of infrastructure and efficient monitoring tools be required by both providers and consumers. Cloud monitoring involves the handling of large amounts of data generated by hundreds or thousands of virtual and physical resources. These resources require distributed monitoring systems in order to be properly monitored and avoid overloaded scenarios. Many of the cloud resources need specific monitoring services that can be different, these differences impact in how they should be distributed by the monitoring systems in order to keep load balancing. In this paper we propose a distribution scheme for cloud monitoring systems, where a set of collectors extract the information of cloud resources in order to reduce the response time when obtaining a global state view of the cloud. Furthermore, we propose a scheme called Policy Aware Allocation (PAA) for load balancing in collectors, where the needs of monitoring for each resource are considered when allocating cloud resources into collectors. The propose schemes were implemented in a distributed monitoring platform. We tested the distribution scheme using different number of collectors and the experiments revealed a reduction in response time when the monitored resources are distributed. For load balancing, we compared our PAA with a standard round robin method, our proposal showed the best results improving load balancing in distributed monitoring systems even when failures occur.