Configuring VMware Server 2.0 Host-Wide Settings
When creating virtual machines within VMware Server 2.0 there is a tendency to focus primarily on the configuration settings specific to each individual virtual machine. One area of configuration which it is important not to overlook when administering a VMware Server based virtualization infrastructure involves the host-wide configuration settings which impact all virtual machines. VMware Server 2.0 provides a number of host-wide configuration settings, an understanding of which is vital to the operation and overall performance of both the virtual machines and host system itself.
Host-wide Memory Settings
Inherent in the concept of virtualization is the fact that a virtual machine must share the physical memory of the host computer system with both other virtual machines and the host operating system. When each VMware Server virtual machine is created, the amount of memory allocated to the particular machine is specified during the initial configuration process and may subsequently be modified using the management tools provided. As such, the host is responsible for allocating memory and will allocate memory based on the requirements of the virtual machines in conjunction with the needs of the host operating system and any other host based processes that happen to be running at any particular time. The key issue with this scenario, however, is that VMware Server is free to request as much memory as necessary to run virtual machines. Without some form of cap on the amount of memory that can requested, there is a danger that so much memory will be requested by virtual machines that there will be insufficient memory for the host operating system and other applications to function, thereby degrading the overall performance of the host, and inevitably also the virtual machines.
In order to address this potential problem, VMware Server allows the configuration of a host-wide VMware Server reserved memory limit. This limit specifies the maximum amount of memory VMware Server can request for running virtual machines. Arriving at the correct setting is something of a balancing act and depends on both the physical memory capacity of the host and the requirements of the virtual machines. When more host memory is requested for virtual machines than is allowed by the limit the memory the virtual machines will perform slowly or fail to power on. If the reserve limit is set too high, there is the danger that the host performance will be impacted if virtual machines use the entire reserved memory. When the memory limit is higher than the current usage, however, the unused portion of the reservation is still available for use by the host and other applications.
