Throughput

Throughput is defined as the quantity of error-free data that is transmitted per unit of time. Throughput is often defined for a specific connection or session, but in some cases the total throughput of a network is specified. Network novices consistently misuse the words throughput and bandwidth. Remember, bandwidth means capacity and is generally fixed. Throughput is an assessment of the amount of data that can be transmitted per unit of time. You measure throughput, which can vary depending on network performance characteristics and how you make the measurement. Bandwidth is a given.

Ideally, throughput should be the same as capacity. However, this is not the case on real networks. Capacity depends on the physical layer technologies in use. The capacity of a network should be adequate to handle the offered load, even when there are peaks in network traffic. (Offered load is the data that all nodes have to send at a particular moment in time.) Theoretically, throughput should increase as offered load increases, up to a maximum of the full capacity of the network. However, network throughput depends on the access method (for example, token passing or carrier sensing), the load on the network, and the error rate.

Figure 2-1 shows the ideal situation, where throughput increases linearly with the offered load, and the real world, where actual throughput tapers off as the offered load reaches a certain maximum.

Figure 2-1. Offered Load and Throughput

Figure 2-1. Offered Load and Throughput

Offer«! Load

Throughput of Internetworking Devices

Some customers specify throughput goals in terms of the number of packets per second (PPS) an internetworking device must process. (In the case of an ATM device, the goal is cells per second, or CPS.) The throughput for an internetworking device is the maximum rate at which the device can forward packets without dropping any packets.

Most internetworking vendors publish PPS ratings for their products, based on their own tests and independent tests. To test an internetworking device, engineers place the device between traffic generators and a traffic checker. The traffic generators send packets ranging in size from 64 bytes to 1518 bytes for Ethernet. By running multiple generators, the investigation can test devices with multiple ports.

The generators send bursts of traffic through the device at an initial rate that is half of what is theoretically possible for test conditions. If all packets are received, the rate is increased. If all packets are not received, the rate is decreased. This process is repeated until the highest rate at which packets can be forwarded without loss is determined. PPS values for small frames are much higher than PPS values for large frames, so be sure you understand which value you are looking at when reading vendor test results for an internetworking device.

Many internetworking devices can forward packets at the theoretical maximum, which is also called wire speed. The theoretical maximum is calculated by dividing bandwidth by packet size, including any headers, preambles, and interframe gaps. Table 2-1 shows the theoretical maximum PPS for one 100-Mbps Ethernet stream, based on frame size.

Table 2-1. Theoretical Maximum Packets per Second (PPS)
Frame Size (in Bytes)	100-Mbps Ethernet Maximum PPS
64	148,800
128	84,450
256	45,280
512	23,490
768	15,860
1024	11,970
1280	9610
1518	8120

To rate the PPS value for a multiport device, testers send multiple streams of data through the device, to multiple output ports. The extreme numbers that you

Cisco's Catalyst 6500 switch—come from measurements made with multiple Gigabit Ethernet data flows, each using 64-byte packets. To understand the maximum number of packets per second on a single Gigabit Ethernet, simply multiply the numbers in Table 2-1 by 10.

Continue reading here: Application Layer Throughput

Was this article helpful?

0 0