Industry experts agree there are few individual applications that require all the processing power available in mid-range and high-end server systems. In most cases, the value of such a server is that it is powerful enough to host several enterprise applications on the same hardware.
Firms that follow this strategy say it reduces operating costs because it simplifies the management of those applications. They may also find this approach improves efficiency because it lets them increase the average utilisation of their server hardware.
Though these ideas are straightforward, firms have found that managing the peak loads of several different applications running on the same server can be problematic. The IT staff will have a clear view of which application should take priority in this situation, but some server management tools are better than others at enabling the staff to enforce such policies.
Without comprehensive tools, the performance of various applications could become unpredictable when two applications both need to use much of the server's RAM or processing power at the same time.
The simplest solution is to use management tools that allow server hardware to be partitioned into separately managed units. Partitioning is also called workload management or logical partitioning. In a basic configuration, each partition is allocated a number of processors, and a percentage of the server's RAM and other resources. Each partition is then configured to run its own operating system and application. When it is known that two or more applications do not contend for the same resources, those applications would probably be hosted in a single partition.
For example, a server might be divided into two partitions, one hosting the production environment and the other hosting a test and development environment. Bringing the test and development environment into production is then likely to be much easier than if it was hosted on a separate - and possibly not identical - system.
While this basic configuration protects applications from being adversely affected by other applications running in other partitions, it does not help to make the most efficient use of the hardware. To do this, the partitioning tools must allow resources such as processors or RAM to be moved between partitions. At the moment, most partitioning tools enable these resources to be re-assigned without stopping applications, and the requirement of many customers is for partitioning tools that can automatically reallocate resources depending on resource utilisation and predefined policies.
Some tools also allow a partition to be completely switched off so that hardware can be replaced; and in some cases different partitions can be comprised of processors of different speeds. Clearly, the flexibility provided by different vendors' partitioning tools is a key differentiator when choosing a supplier. Given that many chipsets for the Itanium 2 support the Numa architecture, we expect the new chip to spur a growth in sales of mid-range systems. The availability and quality of the partitioning tools will become one of the most important factors for firms selecting mid-range and high-end server hardware.
Have your say: contact IT Week











reader comments