When designing an enterprise network, there is only one step between under- and over-provisioning of network resources. However, for many network architects, they worry more about not having enough network resources than about having too much provisioning.

This concern can lead to network performance and throughput far exceeding requirements. This little-discussed practice of network overprovisioning wastes money that could be better spent elsewhere.

Because networks typically last three to seven years, and upgrades can be complex, disruptive, and costly, network professionals should purchase and deploy more network resources—slightly more than required for the entire network lifecycle.

Network configuration and capacity planning

Designing an enterprise network requires some research and buy-in from key business leaders, and follow these steps to avoid overprovisioning.

Network capacity planning techniques include the following 4 steps:

• Review network baselines
• Review the IT roadmap
• Understanding Network Hardware Limitations
• Determine business pivot

1. Review network baselines

When an enterprise plans to upgrade an existing network, a network capacity planning process should be initiated. Network professionals should review network monitoring tools using traditional Simple Network Management Protocol (SNMP) or modern network analysis platforms such as AI for IT operations (AIOps).

These tools monitor network component CPU, memory, and usage levels and graph network activity over time. They contain rich historical network usage and performance data that can serve as the basis for network capacity decisions.

If historical data is available (including months or even years), network architects can calculate the average increase in network usage and performance over time. Network teams can use this information to predict future performance needs of the network.

Network professionals should also use historical information to review network bandwidth, especially at the Internet and WAN edge, to gauge whether throughput performance will reach a point that hinders business operations.

Networking professionals should also use this time to look for other Internet and WAN connectivity options. Many emerging technologies are available at reasonable prices (eg 5G) and, in many cases, multi-year contracts for connectivity services with discounted prices. It is best to calculate the throughput required today and in the months and years to come.

2. Review the IT roadmap

While historical data is useful, teams should also consider future technology projects and their impact on network performance needs. Common IT projects such as cloud migrations, IoT deployments, and big data analytics can lead to dramatic changes in network throughput requirements that increase, decrease, or alter the flow of data across a company's network.

For example, if the IT roadmap shows that applications and data currently managed on-premises will eventually move to the public cloud, the client and server data flow will change. These data flows will no longer be routed to the on-premises data center. Instead, they will access public cloud services over the public internet or direct WAN links.

This roadmap change will dramatically alter the performance and throughput needs of campus routers and switches. This means that as the IT infrastructure changes, the location where teams provide network resources may also change.

3. Understand network hardware limitations

One way to understand hardware and software limitations, including maximum network throughput, is to refer to the network vendor's datasheet, which outlines all features, limitations, and upgrade possibilities.

For example, chassis-based network switches are often able to upgrade their core processors over time to increase backplane throughput. However, the chassis itself has the absolute maximum throughput it can handle, regardless of the processor upgrade. Network professionals should be aware of these limitations to make the most cost-effective decisions when an upgrade is required.

Likewise, switch stacks have become a popular option for the access and distribution layers of enterprise networks. Unlike chassis-based switch hardware, however, with stacking, backplane capacity is permanent and cannot be upgraded. Network architects should keep in mind that when additional switch hardware needs to be added to the stack, each added switch will consume more throughput and will eventually reach the limits of the stack's backplane.

4. Determine the business pivot

The fourth and often most difficult part to review when doing network capacity planning is measuring the stability of the business. Network professionals should adapt their network management strategies to their business needs. Ideally, if the business strategy has been determined, this means no change in network requirements. In this case, the team should not allocate too many resources.

On the other hand, if the business strategy is highly likely to change, this creates resource allocation risks in the network capacity planning process. The pivot of a network strategy may require more or less network resources. Therefore, business leaders need to determine how much over-provisioning risk they can afford. They need to anticipate unforeseen demands, or they need to take the risk of wasting resources - if the business strategy never changes.

This is where a what-if analysis should be performed. Businesses need to identify the possibility of change and conduct a cost-benefit analysis of each network-related situation.

Weighing the costs of configuration and capacity planning

Before beginning this process, the network team should predict the cost of researching and planning for future network configuration and capacity. While inherent knowledge of business needs and future outcomes is an important part of network provisioning and capacity planning, network visibility tools make this work easier and less time-consuming.

Of course, network visibility tools also cost money and time to buy and install. The good news, though, is that the cost of setting up monitoring and analysis tools such as SNMP, AIOps, and NetFlow analyzers is worth the investment. In addition to monitoring, these tools can also address network outages and performance issues. Most enterprises are willing to deploy these tools and use them for a variety of purposes, including network provisioning and capacity planning.