Let us assume that the Oracle database server is deployed on a VMware Guest operating system. If the database expected or required performance is sub par, what do we do? In other words what does the above cartoons Move it to a faster server solution really mean?

In the good old days of non-virtualized servers, databases were tuned and/or optimized at the server level itself because, in many cases, a database instance equated to a single non-shared server. One would simply spend X hours of effort to fix the performance problem and, hopefully as a final resort, reluctantly upgrade the server itself. But with todays extremely cheap hardware, it is almost more cost effective to jump straight to the hardware upgrades (sometimes even in a hapless shotgun approach to improve performance) since computers are now much cheaper than technical person hours especially with todays relatively inexpensive offshore outsourcing!

This is further exasperated because in the virtual server world, one does not really have a physical box associated with the database. Instead, we host the database server on a virtual server, so we are at least one additional level of abstraction further removed. In fact, with dynamic virtualization capabilities and products, some system administrators will simply be able to allocate more hardware resources to virtual machines from a centralized pool of excess capacity as they are needed. Furthermore, some modern virtualization products can increasingly do this automatically within some administrator defined business mandated service level agreement (SLA) thresholds.

Thus, we now have the following choices as legitimate ways to improve database performance:

• Move the virtual machine to a larger host server

• Improve the hardware resources on the host server

• Move the virtual machine to a less utilized host server

• Move another virtual machine off that host server

• Allocate more virtual resources to that virtual machine

• Deallocate virtual resources from other virtual machines

• Increase that virtual machines dynamic load balancing weight

Switch from full virtualization to paravirtualization to better utilize the hosts hardware resources (less host OS overhead)

The first two options are much like our choices of old. But look at all these new alternatives. And as time goes on, there will probably be more. In fact, who knows maybe one day Oracles Grid Computing technology stack will equate to and, therefore, directly support the virtual machine as a key grid component. The future seems almost limitless.

Therefore, the Move it to a faster server solution really does not mean the same thing anymore. As with most things in life, it is now a wee bit more complicated. But that is okay because you can just think of it as more job security!

Virtual Ramifications

The key concept to remember at all times while tuning a virtual machine is that either its underlying hardware resources or virtual hardware allocations may well change over time (usually to increase the capacity to meet increasing demands). So you need to make choices that will automatically scale forward. For once, you may want to error on the side of optimism!

Assuming your virtual host server has been correctly optimized (refer back to the prior chapter), then defining an appropriate virtual machine configuration and finally optimizing its guest operating system is the next logical step. The first part is relatively straightforward and the second part is really no different than tuning a stand-alone database server. You need to make sure that you account for correctly setting configuration files, operating system parameters and database parameters as you always have. The only difference is that CPU, memory and I/O assumptions need to be made more generically since the true hardware has been abstracted and may change over time.

This is an excerpt from Oracle on VMWare: Expert tips for database virtualization by Rampant TechPress.