Performance Management

ICT infrastructure components each have unique performance properties and characteristics, meaning they behave differently in an operational environment. Effective performance management requires a deep understanding of these differences. The Capacity Management process plays a crucial role in overseeing performance management, and any actions taken by Operations should be coordinated with, or directed by, Capacity Management to ensure optimal results.

Understanding Performance Behaviours of ICT Components

When ICT elements near their saturation points, they exhibit different behaviours. For example, a processor operates at full capacity or not at all—there’s no “half throttle.” A system resource monitor indicating a 50% load on a processor means half of the instructions are active program instructions, while the other half are “NOP” (No Operation) instructions. This leads to 100% processor load, but if the demand exceeds 100%, there will be a queue of instructions waiting, indicating a shortage of processing power.

In some cases, a processor load of 90% may not be cause for concern, as every instruction is still executed at full speed. However, a poorly written application could cause a processor to enter an infinite loop, continuously processing a small number of instructions. This loop would drive the processor load to 100%, potentially affecting other programs running concurrently on the same processor. When this happens, Operations staff need to take action, such as terminating the offending process and analysing the memory space to identify the issue.

Managing Memory Saturation and Performance

As system memory nears saturation, all activities may experience delays, particularly because much of the memory is virtual—an imitation of internal memory stored on a slower, less expensive medium like a hard disk. The more memory demand increases, the more noticeable the drawbacks of virtual memory become, primarily in terms of speed. However, several techniques can mitigate these speed penalties:

• Configuration Tuning: Adjusting the configuration of virtual memory can significantly improve performance.
• Caching: Reserving memory regions for frequently accessed data can speed up operations. This cache can also be positioned between the processor and internal memory to accelerate data retrieval.
• Disk Striping: Combining multiple hard disks into a single logical unit and distributing read/write actions across them can enhance data transfer speed, improving virtual memory performance—this is commonly used in RAID arrays.
• Compression: If enough CPU power is available, compressing data can reduce transfer times, as decompressing data may take less time than transferring it uncompressed.

The Law of Diminishing Returns in Performance Tuning

When tuning a system, the initial efforts may yield significant performance gains. However, subsequent tuning efforts tend to produce progressively smaller improvements. This phenomenon is known as the law of diminishing returns, where each tuning effort recovers a smaller portion of the system’s unused performance potential.

Importance of Understanding Component Functionality

To achieve optimal performance, it’s essential to have a solid understanding of the functionality of each component (MO) within the ICT infrastructure. Out-of-the-box configurations are generally reasonable, but specific requirements may necessitate parameter adjustments. Over time, experience, changes in infrastructure, or shifts in usage can reveal opportunities for improvement. Performance tuning adjusts these parameters to progress from acceptable to optimal performance.

Avoiding Infrastructure Bottlenecks

One common pitfall in performance tuning is sub-optimisation, where efforts focus too narrowly on optimising one or two components at the expense of the overall system. This approach can be counterproductive, as alleviating one bottleneck often exposes the next in the sequence. ICT infrastructures inherently consist of a series of bottlenecks, and the end-to-end performance is constrained by the slowest link in the chain.

To avoid sub-optimisation, it’s crucial to consider the overall infrastructure, taking an end-to-end perspective to ensure that improvements in one area do not create new bottlenecks elsewhere.

Optimise Your ICT Infrastructure with Pink Elephant

At Pink Elephant, we specialise in helping organisations optimise their ICT infrastructure to ensure top performance across all components. Our expert team can guide you through the process of performance tuning, bottleneck management, and capacity planning to enhance the efficiency and reliability of your IT services.