Indeed, virtualization will eventually force a move toward a holistic, rather than a server-by-server, management perspective.  As the application becomes further detached from its previously dedicated hardware infrastructure, an older style of element-oriented infrastructure management will become increasingly irrelevant.  Guaranteeing acceptable application service levels will rest on the ability to know at any given point in time exactly what infrastructure components - system resources, network interfaces, WAN links, switch ports, authentication processes, storage resources, etc. - are supporting each application.  It will mean monitoring the performance of that entire delivery chain no matter how often it changes, maintaining an uninterrupted performance picture and using the associated data to preempt or quickly pinpoint the source of service failures.  And the growing number of companies that charge the cost of IT services directly back to their business units will need increasingly sophisticated and agile usage monitoring to adjust to the virtualized aspects of the environment.

If you're a long way from service management nirvana today, are you forced to resort to guerilla tactics? No - you can plot an interim course by applying a service-oriented approach specifically to the server domain.  Though not yet enabling the ultimate cross-domain reach, such an approach represents a big leap toward end-to-end service management, while also yielding immediate efficiency gains, service quality improvements and more risk-free use of server virtualization.

A service-oriented server management platform must have several key qualities:

• It must address the diversity of your environment - in other words, provide a unified picture of utilization levels across all the servers in the environment, and the ability to group them by line of business (as well as other criteria).  Delivering a single pane of glass on the whole server farm means the ability to collect and normalize data from any kind of system - regardless of the mix of Unix, Linux, or Microsoft OS - and preferably without requiring the deployment of new agents.

• It must enable you to not only monitor the performance of VMs, but also to associate the VMs with the physical hosts, and track reallocation of physical resources in near real time.  To tie VM resources to the bigger picture of service quality, the management platform must be able to continually associate a VM with the services it supports throughout the ongoing move-add-change cycle.

• It must have a reporting and navigation interface that allows a high-level service quality view of the businesses' server groups, and drill down as needed to minute layers of technical detail for troubleshooting.

• It must allow you to define performance thresholds and trigger alerts when system performance is trending in a dangerous direction, and ideally provide longer-term trend visibility to predict when resource allocations will not support expected usage levels.

• It must be open enough to readily integrate with other components in the environment and would ideally include out-of-the-box integration with the commonly used virtualization products.

When evaluating such a platform, you should also:

• Look for built-in tools and efficient presentation of capacity information that will help you easily select candidates for virtualization and identify suitable hosts.

• Future-proof your investment by choosing a product that can extend coverage to the network and storage domains of the infrastructure, and incorporate traffic flow and resource mapping data when your company is ready to take the next step toward service management nirvana.

With this system management approach, you could show line-of-business managers a snapshot of utilization levels across businesses servers, immediately highlighting spare capacity.  You can demonstrate the ability to instantly detect when a system is approaching a danger zone and take quick action to preempt outages.  You would also be able to demonstrate an understanding of how applications are using resources today, what the usage growth rates have been and are projected to be and whether or not the business really requires new hardware investments.  If you could hand managers that kind of information about the non-virtualized environment, most would have greater confidence that you could also manage through the risks of virtualization.  And maybe you'd even become anointed as a hero for championing service management.

Infovista