Limitations of Existing VMware Monitoring Solutions
Many solutions for VMware monitoring rely on the vCenter component of the VMware infrastructure to report on the performance of the host and the virtual guests. Since not all infrastructures may have the VMware vCenter available, the reliance on vCenter limits the utility of this approach. Likewise, the granularity and frequency of the monitoring are limited by the capabilities of VMware vCenter. VMware APIs or SNMP are some of the other ways of collecting performance information on the host and the guest operating systems.
These solutions provide an idea of the degree of loading on the host operating system as well as a comparison of the degree of loading that is caused by each of the guest operating systems. However, this only addresses part of the requirements we had outlined for a monitoring solution for virtual solutions. VMware APIs or SNMP interfaces do not provide any information about the resource constraints within each operating system. Hence, a monitoring solution that uses the VMware APIs alone will not be able to determine bottlenecks within a VM guest operating system. For example, consider a case where the space allocated to a VM guest is partitioned across two logical disks. If one of the disks were to become full, monitoring from the ESX server will not be able to report this problem situation, which could be adversely impacting the performance seen by the guest applications.
Many of the existing virtualization monitoring solutions are also not tailored to address the unique requirements of virtual desktop environments. Finally, most of these solutions view the VMware server infrastructure as a silo that needs to be monitored, but they do not consider the inter-dependencies between the physical server and VM guest operating systems and between the applications hosted on these operating systems on the same server. Correlation of performance and usage information across the applications involved in supporting a business service, and identifying the root-cause of a problem by analyzing the virtual machine dependencies of each application involved in supporting the business service is a key value-addition for any monitoring solution for virtual environments.