Virtualisation in data centres, storage systems, and even personal computers has become a major industry trend, yet the process of introducing virtualisation into any data centre can be a daunting task. More importantly, the move to virtualisation must be accomplished in ways that enhance, rather than reduce, the operational flexibility of IT.
Here I will identify seven key things that you must know if your business is to be successful in virtualising your mission-critical applications. Leveraging the right tools goes a long way in ensuring success with your virtualisation environment.
1. The first step in any virtualisation project is inventory
An effective inventory must include the following components:
- Physical server hardware composition;
- Server workload, and installed software and services;
- OS and application licenses;
- Performance baseline leveraging common servers;
- Current and anticipated storage requirements;
- Anticipated growth of server/service needs with special consideration for virtual machine sprawl;
- Existing systems management tools and their capabilities.
This inventory step is important so that an understanding of the physical environment can be leveraged into the architecture for deploying the virtual. Only with this information can you make the right decisions that lead to a successful implementation.
2. Pre-virtualisation performance monitoring is critical
Virtual machines, when collocated with others on the same virtual host, must share the available resources on that host. When one virtual machine oversubscribes resources for its processing, it can impact the others on that same host.
Because of this heightened potential for performance degradation, monitoring for performance and looking for machines with extraordinary needs is an important step in ensuring virtualisation success. Although hundreds, if not thousands, of performance metrics are available, only a few are truly necessary:
- Processor time - the amount of time the processor is performing productive work;
- Processor queue length - high queues can show processors are unable to meet demand;
- Total and available memory - tuning memory to be allocated appropriately is important;
- Memory page rate - how much memory is being swapped from RAM to disk;
- Disk queue length - bottlenecks may appear with high queues;
- Threads and context switch rate - high values here mean many processes are vying for processor attention.
3. Consolidation is only one potential goal
As mentioned earlier, some virtual candidates may not fulfil your goals for consolidation due to high resource use. Yet, in environments where consolidation is less important than virtualisation itself, high-use servers may be virtualised.
In this case, a consolidation ratio approaching 1:1 can be an acceptable goal for deployment - in effect locating a single virtual server atop a virtual host. When you take a hard look at the business benefits associated with virtualisation, other unexpected goals may surface as well.
4. Tailor virtual resources to workload needs
Consolidating virtual workloads onto physical hosts involves attuning assigned physical resources to the needs of the virtual workload. For technical reasons, measuring resource needs from within the virtual machine does not accurately portray those needs from a physical standpoint.
Tools that look at individual processes and behaviours within hosted virtual machines and compare that information with data seen at the virtual host layer are essential. The right tools analyse current and historical data to illuminate where resources might be incorrectly assigned.