At adjustment of a network it is necessary to distinguish nominal and effective throughputs of the protocol. Nominal throughput usually is understood as the bit speed of data transmission supported on an interval of one package transfer. Effective throughput of the protocol is an average speed of the user data transfer, that is the data containing in a floor of data each package. Generally effective throughput of the protocol will be below nominal because of presence in a package of the service information, and also because of pauses between transfer of separate packages.

Let's consider more in detail a difference between nominal and effective throughputs on an example of Ethernet protocol.

Nominal throughput of Ethernet makes 10 Mb/sec that means, that bats inside of the staff are transferred with an interval in 0.1 usec. The staff consists from 8 bytes of a preamble, 14 bytes of the service information - heading, 46 bytes of the user data and 4 bytes of the control sum, all - 72 bytes or 576 bats. At nominal throughput of 10 Mb/sec. Time of transfer of one staff of the minimal length makes 57.6 usec.

Under the standard between the staff the technological pause in 9.6 usec should be maintained. Therefore the period of recurrence of the staff makes 57.6 + 9.6 = 67.2 usec. From here effective throughput of protocol Ethernet at use of the the minimal length staff makes 46 x 8/67.2 = 5.48 Mb/sec.

Real throughput on the user data in a network can be only less than 5.48 Mb/sec resulted above value (for the staff of that size). The attitude of a segment real throughput, the channel or the device to its effective throughput refers to as operating ratio (utilization) a segment, the channel or the device accordingly.

Effective throughput essentially differs from nominal throughput of the protocol that speaks about necessity of orientation to effective throughput at a choice of the protocol type for this or that segment of a network. For example, for Ethernet protocol effective throughput makes approximately 70% from nominal, and for protocol FDDI - about 90%.

Throughput of the protocol is often measured and in quantity of the staff transferred in second. It is easy to count up, that for Ethernet protocol this characteristic for the staff of the minimal length makes 14880 Kb/sec. It is clear, that at measurement of throughput in the staff in a second, there is no sense to differentiate nominal and effective carrying abilities.

Almost all protocols of a channel level of local networks support one fixed nominal throughput: Ethernet - 10 Mb/sec, Token Ring - 16 Mb/sec (4 Mb/sec can be supported for compatibility with the old equipment), FDDI, FastEthernet and 100VG-AnyLAN - 100 Mb/sec. Only protocol ATM can work with various nominal bit speeds - 25, 155 and 622 Mb/sec though transition from one speed to another demands replacement of network adapters or interfaces of switches or routers.

Therefore, if for improvement of a network work we wish to vary nominal throughput of the protocol for this purpose we will need to replace one protocol with another - a measure possible, but demanding significant material and physical expenses.

Time of access by area is defined both logic of the protocol and a degree of a network congestion. Over local networks while are dominated with the divided environments of the data transmission demanding performance of certain procedure for reception of the right staff transfer. In Ethernet and FastEthernet protocols the algorithm of casual access with detection of CSMA/CD collisions, and in protocols TokenRing and FDDI - the algorithm based on determined transfer token of access is used. The new standard 100VG-AnyLAN uses algorithm of access DemandPriority at which the decision on granting access is accepted by the central element - the concentrator.

Time of access area develops of nominal time of access and a waiting time of access. Nominal time of access is defined as time of access to not loaded environment when the unit does not compete to other units. Nominal time of access to the free environment of TokenRing and FDDI protocols in 5 - 10 times exceeds corresponding time of Ethernet protocol as in free network Ethernet the station practically instantly gets access, and to networks TokenRing it should wait arrival of a access marker.

The method of Ethernet protocol access for which sharp growth of a waiting time begins already at sizes of operating ratio in 30% - 50% is most sensitive to congestion of environment. Therefore for normal work of a network it is not recommended to load segments Ethernet over 30%. Even if average value of operating ratio is in norm, but there are the peak values exceeding 60% it is the certificate of that the network works abnormally and demands carrying out of additional researches.

Networks TokenRing and FDDI can be maintained and at great values of operating ratio - up to 60%, and sometimes and up to 80%. Company Hewlett-Packard advancing on the market technology 100VG-AnyLAN, considers, that these networks can normally work and at loading in 95%.