Bandwidth monitoring made easy with PRTG

Bandwidth was originally measured in bits per second and expressed as bps. However, today’s networks typically have much higher bandwidth than can be comfortably expressed by using such small units. Now it is common to see higher numbers that are denoted with metric prefixes, such as Mbps, (megabits per second), Gbps (gigabits per second), or Tbps (terabits per second).

K = kilo = 1,000 bits

M = mega = 1,000 kilo = 1,000,000 bits

G = giga = 1,000 mega = 1,000,000,000 bits

T = tera = 1,000 giga = 1,000,000,000,000 bits

After terabit, there are petabit, exabit, zettabit, and yottabit, each representing an additional power of 10.

Bandwidth can also be expressed as bytes per second. This is commonly denoted with a capital B. For example, 10 megabytes per second would be expressed as 10 MB/s or 10 MBps.

One byte is eight bits.

Thus, 10 MB/s = 80 Mb/s.

The same metric prefixes can be used with bytes as with bits. Thus, 1 TB/s is one terabyte per second.

You usually measure bandwidth using software or firmware and a network interface. Common bandwidth measuring tools are, for example:

• Test TCP utility (TTCP). This tool measures throughput on an IP network between two hosts. One host is the receiver and the other host is the sender. Each side displays the number of bytes transmitted and the time that each packet needs to complete the one-way trip.
• Paessler PRTG. This software provides a graphical interface and charts for measuring bandwidth trends over longer periods of time, and can measure traffic between different interfaces.

Typically, to measure bandwidth, the total amount of traffic sent and received over a specific period of time is counted. The resulting measurements are then expressed as a per-second number.

Another method of measuring bandwidth is to transfer a file or several files of known size and then count how long the transfer takes. The result is converted into bps by dividing the size of the files by the amount of time the transfer required. Most internet speed tests use this method to calculate the connection speed of a user’s computer to the internet.

While there is no way to measure the total available bandwidth, there are many ways to define measured bandwidth, depending on the need.

Theoretical maximum

The highest transmission rate under ideal circumstances. The theoretical maximum transfer rate cannot be achieved in actual installations and is usually only used for comparison purposes, for example, to determine how well a connection is functioning compared to its theoretical maximum potential.

Effective bandwidth 

The highest reliable transmission rate. It is always lower than the theoretical maximum and sometimes considered the best usable bandwidth. The effective bandwidth is necessary for understanding the amount of traffic a connection can support.

Throughput 

The average rate of successful data transfer. It is useful for understanding the typical or usual speed of a connection. Throughput is the size of the transmitted data divided by the time it takes for the transmission to finish. Measured in bytes per second, throughput can be compared to the effective bandwidth and the theoretical maximum as a way of determining how well the connection is performing.

Goodput 

The amount of useful data that is transferred, excluding undesirable data such as packet retransmissions or protocol overhead. Goodput is calculated by dividing the size of the transmitted data by the amount of time the transmission took.

Total transfer method 

Counts all traffic over a set period of time, typically a month. This is most useful for billing based on how much bandwidth is used.

95th percentile method 

To avoid having bandwidth measurements skewed by spikes in usage, carriers often use the 95th percentile method. The idea is to continuously measure bandwidth usage over time and then remove the top 5 percent of use. This is useful for billing based on how much bandwidth is usually used in a set period.

In real-world networks, bandwidth varies over time depending on usage and network connections. As a result, a single bandwidth measurement says very little about the actual bandwidth usage. A series of measurements can be more useful when determining averages or trends.

Speed

There are many ways to think about the flow of data in a network. The speed of a network is defined as the bit rate of the circuit, determined by the physical signal speed of the medium.

Bandwidth

Bandwidth is how much of the physical circuit’s capacity can be used to transmit data and is determined by how much of the network capacity is available based on the connection. While a Gigabit Ethernet network connection would allow for 1 Gbps, the bandwidth available to a computer connected by a Fast Ethernet card would only be 100 Mbps.

Throughput

Throughput is the rate of successful transmission, while bandwidth is a calculation of the amount of data that passes the network interface, regardless of whether the data results in a successful transmission. As such, throughput is always lower than bandwidth.

There are several reasons to measure bandwidth. Low usable bandwidth compared to the theoretical maximum bandwidth may be indicative of network problems, particularly if there are widely different usable bandwidths from different parts of a network that are designed to operate the same.

Additionally, measuring bandwidth is necessary to ensure that any paid connections are living up to their promise. Home users may run an online bandwidth test such as the DSLReports speed test. Corporate connections might be better served by measuring throughput between offices connected by a carrier-leased line connection.

Bandwidth management

To implement proper bandwidth management or QoS controls, one must first understand what bandwidth is used. Once determined, a continuous measurement will ensure that all users get the necessary bandwidth.

Bandwidth throttling

Once you understand bandwidth usage patterns and if specific users or applications degrade network performance for others, you can use tools to limit the amount of bandwidth they use.

Bandwidth maximums

Some types of connections have a maximum defined bandwidth. Actual bandwidth depends on many factors including environment, cabling, and usage, and is usually less than the theoretical maximum.

Bandwidth is most often purchased from telecommunications companies. Most consumer bandwidth is sold as "up to" meaning that the customer may get up to 40 MB/s, but not always have that speed while using the connection.

Speeds may be higher or lower at different times of the day or under different circumstances. Corporate bandwidth is also typically purchased from telecommunications companies. However, many corporate agreements come with contractual performance measures that must be met, including a minimum usable bandwidth, minimum uptime, and other metrics.

Additionally, bandwidth metering may be used to charge for specific usage rather than a full connection. For example, a website owner may pay the website host only for the amount of bandwidth used by that specific website over a period of time, such as a monthly billing period.

Too little bandwidth

While modern protocols are pretty good in not losing any packets, limited bandwidth can still cause some issues:

• operations take too long to complete, resulting in timeouts or process failures like application or database errors or backup malfunctions
• users can notice long lag times between when they do something, like clicking a button, and the response to that action
• for users attempting to make phone calls over a network, such as VoIP, having too little bandwidth results in low-quality calls
• video calls made without the necessary bandwidth will not only result in bad sound quality, but also low quality or jittery video

For internet users, the United States Federal Communications Commission (FCC) recommends a minimum bandwidth of 4 Mbps for adequate performance when streaming a video in HD quality. Many video players can work with less bandwidth by “buffering”, that is, downloading data ahead of when it is actually displayed.

Gamers are often frustrated by limited bandwidth as well. While playing against other players online, players with faster connections see what is happening quicker, and the data about their reactions is transmitted and received faster. The FCC recommends a minimum download speed of 4 Mbps for Online Multiplayer Gaming in HD.

Too much bandwidth

There a few technical issues caused by too much bandwidth. Higher capacity bandwidth, however, typically costs more. Thus, too much bandwidth may not be cost effective.

Latency

Network design and infrastructure can create bandwidth issues as well. Latency measures the delays on a network that may be causing lower throughput or goodput. A low-latency network has short delays, while a high-latency network has longer delays. High latency prevents data from fully using the network’s capabilities, therefore decreasing the bandwidth.

Finding and troubleshooting bandwidth issues helps improve network performance without costly upgrades.

Ping and traceroute

Tools such as Ping and traceroute can help troubleshoot basic issues.

• Pinging a test server, for example, will return information on how quickly data can be sent and received, as well as average round-trip times. High ping times indicate higher latency in the network.
• A traceroute tool can help determine if there are too many individual network connections, or hops, along the connection path. In addition, traceroute returns the time taken by each hop. A longer time on a single hop may point to the source of an issue.

TTCP

TTCP measures the time it takes for data to travel from one network interface to another with a receiver on the other end. This eliminates the return trip from the calculation and may help pinpoint issues quickly. If the measured bandwidth is less than expected, further measurements can isolate the issue.

Bandwidth monitoring software

The bandwidth monitoring tool PRTG can also help troubleshoot bandwidth problems that are not related to design. By measuring bandwidth usage over time, for example, you can detect if specific users or applications sometimes use higher amounts of bandwidth and cause network congestions.

In PRTG, “sensors” are the basic monitoring elements. One sensor usually monitors one measured value in your network, for example the traffic of a switch port, the CPU load of a server, or the free space on a disk drive.

On average, you need about 5-10 sensors per device or one sensor per switch port.