Network capacity planning tutorial

What is Network Capacity Planning?

Networks evolve with a business. Whether you are planning a new LAN or maintaining an existing one, you need to model the data capacity requirements on the business’s needs. So, your starting point for network capacity planning is to document your existing resources, the networks current performance metrics, and assess potential changes in demand.

If you currently document your network manually, you will need to start off by finding network management software to use as a network capacity planning tool to map the network and gather statistics on its performance.

Planning future requirements

The basic outline of core capacity planning tasks is very straightforward. You just need to work out the following three factors:

  1. Current system potential capacity
  2. Current capacity utilization
  3. Bandwidth demand of new equipment/software

If you currently run a network, you may scoff at the simplicity of that outline. However, when performing core capacity planning, it really does pay to keep things as simple as possible. In the baseline phase, you need to consider the load on the network links and equipment that will be needed to support the new requirement. For example, if you are adding on 20 new user endpoints to the network, the segment that will connect those new nodes together and links that will carry traffic between those terminals and the servers that hold the storage and software that those users are expected to use are key. Your firewall and load balancer capacity is not relevant in this scenario.

So, capacity planning only needs to concern those resources that will be directly involved in serving the new requirement. You will be expected to add on new applications or equipment throughout your career managing your company’s network. Therefore, it is better to get a baseline for the whole system upfront, so that you are ready to implement any new requirement that the business comes up with. Those system statuses need to be updated regularly so that future capacity planning exercises aren’t based on out-of-date network metrics.

Establishing a baseline

You will need to take into account the following factors when gathering information about your current system:

  1. Network equipment
  2. End-user equipment
  3. On-premises servers
  4. Offsite servers
  5. Outsourced services
  6. Existing applications
  7. Subnets
  8. Remote access requirements
  9. VM demands
  10. External traffic demand

Getting your baseline information involves measuring the existing network’s performance. This involves tracking data from both your network devices and your link capacity at the moment. Understanding which business critical applications are delivered to which users is also a benefit because this can help you re-organize subnetworks to reduce pressure on trunk links. You can also reorganize scheduled tasks to reduce demands on important links and free up bandwidth for user-facing applications during business hours.

Planning for New Networks

Creating a network from scratch to support a fully-operational new business with no historic network performance is a tough job. Fortunately, it is also very rare. No business starts up fully formed. New businesses need to perform market research, run trials of products and services and assemble a team. So you will have a period when you can try your new network and check on the performance of each application and business service as you test it on the server. So, you never really will be required to create a new network that will be live on day one. Capacity planning is a recursive process that takes place as the business sets up.

Monitoring tools

Data gathering on networks is a complicated task that requires views on a wide range of information types that originate from a large number of sources. It is impossible to keep track of all the variables that can impact network performance without an automated monitoring tool. So, during your baselining task, you need to use an efficient monitoring suite to gather all of the information that establishes an acceptable level of service.

Logging equipment

Your first task, when establishing a baseline, is to create an inventory of all of your equipment. For network capacity planning you need to focus on your network devices, such as switches and bridges. For later phases in the traffic capacity planning exercise, you will also need to list your end user devices and servers. If you have wifi routers in your network, you will also need to record the data throughput rate of each of them. If you use Software-as-a-Service online (SaaS) or employ cloud storage, you need to include those offsite resources as well.

Some network monitoring tools are able to cover entire WANs, so if you want to monitor many sites as an integrated network, you should be sure to choose a monitor that has multi-site capabilities.

You specifically need to know the total throughput capacity of each piece of network equipment. You should already know the bandwidth capacity of your cable type. It is not unusual for networks to include a mix of cable types. The upstream and downstream bandwidth availability of your internet service and the capacity of interfaces to external networks should also be logged. The throughput capacity of your firewalls, gateways, load balancers, and proxy servers also needs to be recorded.

With this data gathering phase, you fulfill the first information requirement of your capacity planning exercise, which is the potential maximum capacity of your network.

Map the network

Getting a visual representation of the network is very important. You may have a good equipment inventory tool that allows you to sort data and focus on each piece of equipment. However, there is nothing like a map to enable you to see instantly which switches are serving more links than the others. A visual representation also makes it easier to see how many links a typical conversation has to cross. The network mapping tool that you use should be able to display capacity utilization through color-coding so you can spot congestion instantly.

If your business is web-based, your network traffic analyzer should specialize in tracking demand into your network from external sources. If you rely heavily on cloud-based services, your monitor should be able to track connections to those remote servers. Similarly, your bandwidth analyzer should be able to monitor internet links between sites if you run a WAN.

Gather capacity usage data over time

When you monitor your network, you will get to see which hours of the day, and which days of the month experience peak traffic loads or spikes, and when traffic levels are subdued. During regular operations, you should be able to schedule administrative tasks to run outside of business hours in order to avoid competing with user traffic at peak times. So, before you add on new services, you should have your network working efficiently. When your capacity planning study begins, you need to track the capacity utilization over several working days so that you can get a view of your peak network utilizations over the resources that will serve the new demand.

Get a record of high, low, and average throughput on the links and equipment that are due to work for the new requirements. You need to know whether the peak demand is close to typical or only occurs on a limited period on any given day.

You also need to measure the packet loss rate. You should record this figure as a percentage. See whether the packet loss rate increases during high demand periods. Apply the packet loss rate to your bandwidth availability by deducting the calculated percentage from your figure of available bandwidth. Alternatively, you can choose to add that percentage to the new bandwidth requirements.

Now you have the second input to your capacity planning. That is, the current data throughput of your network.

Estimate new bandwidth demand

Getting a figure for bandwidth requirements for any new software that you intend to install is pretty easy. The providers of most software that is to be delivered across a network give information on the bandwidth that their software uses in its Systems Requirements statement. In some cases, this figure is expressed as a range. If that range is wide, the information won’t be much help to you. Usually, the bandwidth requirements express the throughput expected to serve one user, so if you are buying several licenses, you should plan to have the extra capacity to cater to the given bandwidth requirement multiplied by the number of licenses.

If the software company does not provide a bandwidth requirement, ask for a trial copy and run it over the network as a single instance to log its throughput. The commissioning manager in your company who requested the installation of this software will probably require an assessment period. That run through will enable you to gather bandwidth requirement data.

Your monitoring tool should enable you to track the traffic of a specific piece of software. If it doesn’t, you should be able to track protocol activity. If your new software is replacing a package that is already in use on the network, both the new and old software will probably use the same protocol or port number. In that case, you would need to record that traffic outside of normal business hours so that you don’t confuse the bandwidth utilization used by both systems. Another option is to employ QoS tagging so that you can run the new software test while the old system is still in operation. Once you have a standard traffic pattern for one user, multiply that figure by the maximum number of users that will be allowed to use that software simultaneously.

If your capacity planning exercise examines the effects of adding on new user devices to the network, record the amount of router traffic to and from an existing node that serves the same work roles that will use the new devices. Multiply those traffic figures by the number of endpoints that you expect to add to the network.

The figures you acquire with this phase of the study give you the third element of data that the capacity planning task requires — new bandwidth demand.

Basic network capacity planning

Your network performance monitor and your bandwidth analyzer should have provided you with the input data for your capacity planning inquiry. Hopefully, you also have a network administration tool that supports analysis functions.

The capacity test is a simple exercise of adding the new bandwidth requirements to your current peak and average bandwidth usage. Next, look at the maximum capacity of each piece of equipment and each network link that the traffic will pass through. You should know this route thanks to your network map. Look at the network element that has the lowest capacity. Don’t forget to adjust your figures for the packet loss rate. If your new demand is below that level, you are all clear.

If your tightest bottleneck can’t cope with the new traffic, look at other elements in the path in order of increasing capacity to see which of those also could not cope with the new traffic demand.

You may find that you have more than one element that needs to be replaced to keep the new service operating at an acceptable speed.

Capacity refinements

The discovery that your network cannot cope with the new capacity demand means that you need to make some decisions and you will probably need to include others in your decision making. You will need to present options to other managers in the business and so a network administration tool that has reporting functions will really help you in this task.

In any infrastructure capacity decision you always have three options:

  1. Buy new resources.
  2. Reorganize the network to add capacity to the path of this requirement.
  3. Cancel the new acquisition.

The initial capacity test showed you a list of equipment that could not cope with extra bandwidth demand. Get quotes for replacements of each element with new equipment that has greater capacity. This will give you the cost of the first option on the above list: buying new resources. One advantage is that you don’t have to throw your existing equipment away. It can be moved to other parts of the network to provide spare capacity that will cater to future extra demand. Every network administrator should plan for a continual increase in network bandwidth.

No-cost capacity options

You will need more information about your network to support this decision-making process. The no-cost option of reallocating underutilized resources is always going to be the choice of the board of directors, so you first need to investigate that possibility before raising their hopes. You can go in three directions to get the most out of your network:

  1. Physically move network equipment.
  2. Reallocate resources virtually.
  3. Implement traffic shaping.

All of these options require levels of analysis that cannot be achieved without network administration tools.

Reassign physical equipment

Option one on the no-cost options list would be nice, but you probably don’t have a spare switch on your network with links that carry almost no traffic. If you do, then you have probably overspent at some point in the past. The traffic throughput of all of your network devices shown on the network map will aid your decision making when looking for spare capacity. A printout of your network map, showing the capacity and utilization of all your links will add a visual element to your presentation when you need to include non-technical managers in the planning process.

Use your network administration tool’s analysis functions to log the maximum load on each switch and router. If there are two that serve less than half of their maximum capacity and one of them has enough spare sockets to take more cables, then you have the option of merging two segments and freeing up a switch. With that extra equipment, you can divide up the overloaded segment and double your capacity to those endpoints.

This is just one example of the steps you can take to physically expand the capacity on one part of your network at the expense of another in order to cater to new traffic demand. The monitoring capabilities and analytical functions of your network administration tool will get you through this complicated task.

Introduce virtualization

Virtual reallocation is possible if you have virtualization applications, such as Hyper-V or VMWare. Sharing a server among endpoints not only reduces hardware costs, but it can also reduce current network traffic volumes. When users log in to the processor remotely, much of the interfacing between applications and storage can be limited to one server or run over one link. This generates less traffic than messages passing back and forth across the whole network between user endpoints and servers.

Virtualization is complicated and difficult to assess, so you need a tool that is dedicated to monitoring the traffic that these systems create. If you don’t already implement virtualization, it is probably better to put off this option for consideration at a later date when you have less pressure on your time.

Implement traffic shaping

If you don’t have spare capacity anywhere on your network and the new requirement is going to overload part of your system, you could propose traffic shaping. The debate between buying new equipment and traffic shaping is one of the main reasons that you need to get senior staff involved in the decision over how to cater to the new requirement. Being able to distill network capacity information into easily-digestible graphs and summaries will ease the process of trade-off. Someone is not going to be happy about the results of this exercise.

If you tracked your network throughput over time and stored that data, your analysis tool will be able to replay that traffic and segment it by protocol and endpoints. So, you can identify which traffic generating tasks can be moved to overnight processing in order to free up capacity. If you have already used up that option to squeeze extra value out of your network budget, then you still have one more traffic shaping option: prioritization. Try queuing algorithms, such as Class-based Quality of Service (CBQoS) to ensure that some traffic gets through faster than others.

Your capacity planning tool will help you see the effects of this move. However, it ultimately means that some users will get slower response times for the applications that they need to use interactively. You should be able to print off reports from your planning tool to illustrate the effects of this policy.

Capacity planning decisions

If the existing services cannot be slowed down to create sufficient bandwidth space for the new requirement, your management team will be faced with the decision of spending money to buy new infrastructure or canceling the new project. These are decisions that are not within the realm of the network administrator. It would help if you prepared all of the information needed to support the decision and provide stakeholders with illustrated scenarios. The reports from your monitoring tool and snapshots generated out of your capacity planning software will enable you to furnish this supporting documentation.

Other capacity issues

When you intend to expand the services of your network, there are peripheral issues to consider. You may have room on your server for new software, but will your server’s CPU be able to handle all of the additional demand for processing? Will your server’s network card still be able to send out and receive in all of the traffic that your existing programs generate and still have bandwidth for these extra requirements?

If you interface heavily with the internet, your firewalls, load balancers, proxy servers, and DDoS mitigation capacity may need to be upgraded. Does your internet service agreement have throughput limits? Will your service provider be able to cope with the extra internet traffic that your system changes might generate?

The network does not operate in isolation, so there will be other hardware considerations to take into account when you plan to expand. In each case, measuring current resource usage, recording resource capacity limits, and estimating new capacity requirements will all have to be applied in a capacity planning phase before you can start the implementation of your new project.

Network capacity planning

No business is static and hopefully, the constant change that your organization experiences is all generated by expansion. You need to allow for traffic growth which will occur naturally as the business grows even if your users don’t demand any new software or hardware. As a network administrator, you need to keep an eye on live data showing traffic volumes because they will rise and the alerts that your system raises to notify you of approaching capacity exhaustion will occur more and more frequently until you add on new network resources.

Resource monitoring and capacity planning are two activities that go hand-in-hand. There are many reasons why you may need to run a capacity planning exercise. However, one constant is that you will always need to have very detailed information on your network’s current activities before you can start to plan for the future.

Network Capacity Planning FAQs

How can AI improve network capacity planning?

The Machine Learning discipline of AI is the most helpful advance in capacity planning technology. Machine Learning works on statistical probability and the more source data an ML system has, the better it can predict future capacity needs.

How is network capacity measured?

Network capacity measurements count the number of bits that pass down a link in each second. As volumes are high, the usual unit of measure is Megabits per second (Mbps). Recently, faster speeds have become so great that they are measured in Gigabits per second (Gbps). A Megabit is one million bits and a Gigabit is one billion bits or 1,000 Megabits.

Image:  Information Technology Industry 4 from Max Pixel. Public domain