Capacity planning is not unique to Information Technology Infrastructure. We see the need for capacity planning in many industries and organizations.
For IT infrastructure, capacity planning involves the need to estimate the demand for infrastructure components such as processing power, storage and network bandwidth. If the infrastructure is hosted centrally in a data center, capacity planning needs to take into account the physical space required to add more infrastructure components in the future.
Processing Power
Processing power depends primarily on the following elements:
The actual processing power that is perceived by an application is dependent on the nature of the application (is the application CPU-intensive or I/O-intensive?, is the application single-threaded or multi-threaded) and many other factors such as the ability of the processor to process multiple threads simultaneously and the ability of multiple processors to work in parallel.
Storage
Storage is a central element of many information systems. Storage is of particular importance to large organizations and to large service providers (examples include: public email providers such as Google Gmail, Yahoo, Microsoft Hotmail etc., search engines such as Google Search, Microsoft Bing, media repositories such as Lexis Nexis etc.)
Storage systems and components come in many different flavours. Some are made for very fast access (think solid state disks) and others are made for slow access and larger capacities (think optical storage).
Capacity planning for storage involves determining the need for fast and slow access to storage, segmenting storage systems into categories and identifying the required capacity for each segment.
Often, capacity utilization increases faster than expected; especially when the usage of an application takes off. So, capacity planning involves allowing for unforeseen usage for storage.
Network Speed and Bandwidth
Modern information systems are designed such that most processing happens in a centralized facility and the processed results are delivered to uses in many locations. The need to transport the processed results to user locations as well the need to ensure that the data transfers between processing components are fast requires that the speed and the bandwidth of the network that connects them are adequate.
The race to provide better speed and bandwidth has been a key factor in the competition between various network service and equipment providers over the past two decades.
Network connectivity may be wired or wireless. Thus far, most connectivity between large systems is based on wired connections (1G, 10G, Fiber etc.). However, consumer connections are increasingly becoming wireless - WiFi, 3G, 4G etc.
Typically, the greater the speed and the bandwidth, the greater the cost. Other factors that influence cost are the quality of support and the amount of redundancy and fault tolerance that is built into the network.
Capacity planning is not an easy task. The uncertainty in how the infrastructure will be used in the future makes it very challenging. There is a need to balance the cost of paying for excess unused capacity to the probability that utilization by applications will exceed the planned capacity.
Cloud computing and virtualization try to pool capacity into structures that are uniformly accessible by applications and thereby potentially increasing the capacity utilization without increasing the risk of exceeding the planned capacity. We will look at cloud computing and its relationship to capacity planning in another article.
Copyright © 2012 Suresh Peram. All rights reserved.
- A manufacturer of auto parts needs to plan the factory capacity meet the expected demand
- A city needs to plan for the capacity of its electricity, water supplies and transportation
- A family or an individual needs to plan for how much living space they would need
- A school needs to plan to accommodate the expected number of students
For IT infrastructure, capacity planning involves the need to estimate the demand for infrastructure components such as processing power, storage and network bandwidth. If the infrastructure is hosted centrally in a data center, capacity planning needs to take into account the physical space required to add more infrastructure components in the future.
Processing Power
Processing power depends primarily on the following elements:
- The raw speed and the architecture of the central processing unit
- The raw speed of the primary memory and the speed of the interconnection between the central processing unit and the primary memory
- The raw speed of the storage unit and the speed of the interconnection between the central processing unit and the storage unit
The actual processing power that is perceived by an application is dependent on the nature of the application (is the application CPU-intensive or I/O-intensive?, is the application single-threaded or multi-threaded) and many other factors such as the ability of the processor to process multiple threads simultaneously and the ability of multiple processors to work in parallel.
Storage
Storage is a central element of many information systems. Storage is of particular importance to large organizations and to large service providers (examples include: public email providers such as Google Gmail, Yahoo, Microsoft Hotmail etc., search engines such as Google Search, Microsoft Bing, media repositories such as Lexis Nexis etc.)
Storage systems and components come in many different flavours. Some are made for very fast access (think solid state disks) and others are made for slow access and larger capacities (think optical storage).
Capacity planning for storage involves determining the need for fast and slow access to storage, segmenting storage systems into categories and identifying the required capacity for each segment.
Often, capacity utilization increases faster than expected; especially when the usage of an application takes off. So, capacity planning involves allowing for unforeseen usage for storage.
Network Speed and Bandwidth
Modern information systems are designed such that most processing happens in a centralized facility and the processed results are delivered to uses in many locations. The need to transport the processed results to user locations as well the need to ensure that the data transfers between processing components are fast requires that the speed and the bandwidth of the network that connects them are adequate.
The race to provide better speed and bandwidth has been a key factor in the competition between various network service and equipment providers over the past two decades.
Network connectivity may be wired or wireless. Thus far, most connectivity between large systems is based on wired connections (1G, 10G, Fiber etc.). However, consumer connections are increasingly becoming wireless - WiFi, 3G, 4G etc.
Typically, the greater the speed and the bandwidth, the greater the cost. Other factors that influence cost are the quality of support and the amount of redundancy and fault tolerance that is built into the network.
Capacity planning is not an easy task. The uncertainty in how the infrastructure will be used in the future makes it very challenging. There is a need to balance the cost of paying for excess unused capacity to the probability that utilization by applications will exceed the planned capacity.
Cloud computing and virtualization try to pool capacity into structures that are uniformly accessible by applications and thereby potentially increasing the capacity utilization without increasing the risk of exceeding the planned capacity. We will look at cloud computing and its relationship to capacity planning in another article.
Copyright © 2012 Suresh Peram. All rights reserved.
