8.1 Token Ring Traffic Modeling


8.1 Token Ring Traffic Modeling

In comparison to Ethernet, the modeling of a Token Ring network can be much more complex. This is because the frame rate depends on the number of nodes in a network, the token holding time per node, the type of wire used for cabling and ring length, and the type of adapter used as a ring interface unit.

The number of nodes and their cabling govern both token propagation time and holding time as a token flows around the ring. The type of adapter used governs the maximum frame rate supported. This rate can vary between vendors as well as within a vendor's product line. For example, Texas Instruments' original MAC code permitted a maximum transmission of 2200 64-Kbyte frames per second. New software from that vendor raised the frame rate to 3300, and a more recent release known as Turbo MAC 2.1 increased it to 4000 frames per second.

The type of cabling and ring length governs the propagation delay associated with the flow of tokens and frames around the ring. Although the data rate around a ring is consistent at either 4 Mbps or 16 Mbps, tokens and frames do not flow instantaneously around the ring and are delayed based upon the distance they must traverse and the type of cabling used. In addition, a slight delay is encountered at each node because the token must be examined to determine its status.

8.1.1 Model Development

In developing a model to determine Token Ring frame rates, let us assume there are N stations on the network. Then, on average, a token will travel N/2 stations until it is grabbed and converted into a frame. Similarly, a frame can be expected to travel N/2 stations until it reaches its destination and another N/2 stations until it returns to the origination station and is reconverted into a token.

As mentioned at the beginning of this chapter, the performance of a Token Ring network typically depends more on the total network cable length than an Ethernet network. Thus, in developing a model of Token Ring network performance, a mechanism is required to equate cable length to the propagation delay of electrons flowing on the cable. To do so, let us start with the well-known velocity of light.

8.1.2 Propagation Delay

In free space, the velocity of light is 186,000 miles per second. In twisted pair cable, the speed of electrons is approximately 62 percent of the velocity of light in free space. Thus, electrons will travel at approximately 186,000 * 0.62, or 115,320 miles per second. Because there are 5280 feet in a mile, this rate is equivalent to 608,889,600 feet per second, or approximately 609 feet per microsecond. Then, to traverse 1000 feet of cable would require 1000/609 * 10 6 , or approximately 1.64 * 10 ˆ’ 6 seconds.

At a Token Ring operating rate of 4 Mbps, the bit duration is 1/4,000,000, or 2.5 * 10 ˆ’ 7 seconds. At a network operating rate of 16 Mbps, the bit duration is 1/16,000,000, or 0.625 * 10 ˆ’ 7 seconds. Because the time required for electrons to traverse 1000 feet of cable is 1.64 * 10 ˆ’ 6 seconds, the cable propagation delay time per 100 feet of cable can be converted into a bit time delay to simplify computations . At a Token Ring operating rate of 4 Mbps, the bit time delay per 1000 feet of cable becomes 1.64 * 10 ˆ’ 6 /2.5 * 10 ˆ’ 7 , or 6.56 bit times. When the Token Ring network operates at 16 Mbps, the bit time delay per 1000 feet of cable becomes 1.64 * 10 ˆ’ 6 /.625 * 10 ˆ’ 7 , or 26.24 bit times.

8.1.3 4-Mbps Model

In developing a Token Ring performance model, let us commence the effort by assuming that the network operates at 4 Mbps. Once we develop this model and use it to perform a series of manual calculations, we will then construct a general model applicable to both 4- and 16-Mbps networks. That model will then be used as a basis for developing a BASIC language program that will be used to generate a comprehensive series of tables contained at the Web URL as well as two tables in this chapter whose entries we examine in detail. The tables contained at the Web URL can be used to facilitate determining the frame flow on a Token Ring network because they enable the frame flow computations to be supplemented by a table lookup process. The tables in this chapter can be considered to represent extracts from the comprehensive table and their use serves as a guide to the use of the comprehensive series of tables.

For the development of a 4-Mbps Token Ring performance model, let us start with the flow of a token as indicated by the following steps in the model development process.

  1. Given a Token Ring network with N stations, a free token travels , on average, N/2 stations until it is grabbed and converted into a frame.

  2. Each station adds a 2.5-bit time delay to examine the token. At a 4-Mbps ring operating rate, a bit time equals 2.5 * 10 ˆ’ 7 seconds. Thus, each station induces a delay of 2.5 * 2.5 * 10 ˆ’ 7 , or 6.25 * 10 ˆ’ 7 seconds.

  3. The token consists of 3 bytes, or 24 bits. The time required for the token to be placed onto the ring is:

  4. The time for the token to be placed onto the ring and flow around half the ring until it is grabbed is the sum of the times of step 2 and step 3. This time then becomes:

  5. Once a token is grabbed, it is converted into a frame. On average, the frame will travel N/2 stations to its destination. A frame containing 64 bytes of information consists of 85 bytes because 21 bytes of overhead, including starting and ending delimiters, source and destination addresses, and other control information must be included in the frame. Thus, the time required to place the frame on the ring becomes:

  6. If the network contains N stations, the frame must traverse N/2 stations, on average, to reach its destination. Thus, the time required for the frame to be placed onto the ring and traverse half the ring becomes:

  7. The total token and frame time from numbers 4 and 6 above is:

    click to expand
  8. Once the frame reaches its destination, it must traverse another N/2 stations, on average, to return to its originating station, which then removes it from the network. When this occurs, the origination station generates a new token onto the network and the previously described process is repeated. The time for the frame to again traverse half the network becomes:

    This time must be added to the time in step 7. Doing so, we obtain:

  9. To consider the effect of propagation delay time as tokens and frames flow in the cable, we must consider the sum of the ring length and twice the sum of all lobe distances. Here, we must double the lobe distances because the token will flow to and from each workstation on the lobe. If we let C equal the number of thousands of feet of cable, we obtain the time in seconds to traverse the ring as:

    which equals:

    where:

    N = number of stations

    C = thousands of feet of cable

8.1.4 Exercising the Model

To illustrate the use of the previously developed Token Ring performance model, let us assume a Token Ring network of 50 stations has 8000 feet of cable. Then, the time for a token and frame to circulate the ring becomes:

Then, in 1 second there will be, on average, 1/2.36 * 10 ˆ’ 4 , or 4237 64-byte information frames that can flow on a Token Ring network containing 50 stations and a total of 8000 feet of cable.

8.1.5 Network Modification

To illustrate the use of the previously developed model in determining the effect of cabling and network stations on the frame rate, let us now consider what happens when the network is reduced in size . Suppose the number of workstations is halved to 25 and the total cable distance reduced to 4000 feet. Then, with N = 25 and C = 4, the time for a token and frame to flow around the ring becomes:

Thus, in 1 second there will be, on average, 1/2.06 * 10 ˆ’ 4 , or 4854 64-byte information frames. As we would intuitively expect, as the number of stations and cable distance decrease, the transmission capacity of the ring increases .

8.1.6 Varying the Frame Size

Now let us examine the effect of transmitting larger information frames. Suppose we transmit 4000-byte information frames. Here, a total of 4021 bytes is required. Thus, the time required for the frame to be placed on the ring becomes:

Then, the total token and frame time becomes:

Again, let us assume the number of stations, N, is 50, while the cabling distance is 8000 feet. Thus, we obtain the token and frame revolution time as follows :

Then, in 1 second there will be 1/81.08 * 10 ˆ’ 4 or 123.3 frames. Because each frame contains 4000 bytes of information, the effective operating rate becomes 123.3 * 4000 * 8, or 3.946 Mbps for a 50-station Token Ring network with 8000 feet of cable using 4000 character information frames. In comparison, a similar Token Ring network using 64-byte information frames would have a frame rate of 4237 frames per second. However, this rate would be equivalent to an information transfer rate of 4237 * 64 * 8, or 2.169 Mbps. Thus, larger frame sizes provide a more efficient data transportation capability.

The preceding computations, which represent the use of a simplified model of a 4-Mbps Token-Ring network, indicate an important concept. That is, the frame length, cabling distance, and number of network stations govern the maximum frame rate that can flow on a Token Ring network. This tells us that when a network becomes saturated due to heavy usage, you should consider breaking larger networks into two or more subnets interconnected by bridges to improve Token Ring network performance. The preceding is a simplified model due to the fact that the model does not include the effect of the flow of network management frames which, when they flow on the network, preclude the transfer of data. For example, every seven seconds, the active monitor transmits an Active Monitor Present frame for which all other stations respond with a Standby Monitor Present frame. Because the frame rate on a Token Ring network will range from over 100 frames per second when the frame length approaches the maximum size frame length on a 4-Mbps network to many thousands of frames per second when the minimum length frame is transmitted, the effect of the Active Monitor Present frames and responding Standby Monitor Present frame every seven seconds is negligible upon network performance. This is because those two frames are relatively short and would result in a maximum of 260 frames every seven seconds, during which approximately 15,000 or more similar length frames could be transported on the network. Thus, the use of a simplified model does not materially affect our model.

8.1.7 Adapter Card Considerations

One of the more interesting aspects of Token Ring frame rates is that the majority of adapter cards from different vendors that use the Texas Instruments chip set support a maximum frame rate of 4000 frames per second. This indicates that a further constraint on the number of nodes and cable length is the adapter cards used in a network. In 1991, Madge Systems introduced an adapter card capable of transmitting approximately 12 thousand 64-byte frames per second. Thus, using that firm's adapter card or other higher-performance adapter cards manufactured by other vendors can significantly improve the performance of a Token Ring network. However, the use of such "high-performance" adapter cards is irrelevant when a network grows in size in terms of the number of network stations and cable distance. In such situations, the capability of high-performance network adapter cards cannot be effectively used.

Now that we have developed and exercised a mathematical model to determine the frame rate on a 4-Mbps Token Ring network, we will use our prior effort to develop a general model for 4- and 16-Mbps networks. In doing so, let us use BASIC language variables so that we can exercise our model through its incorporation into a BASIC language program.

8.1.8 General Model Development

To denote the difference between 4- and 16-Mbps networks, let us use the array variable BITTIME(I). Then, we can assign the value 1/4,000,000 to BITTIME(1) to represent the bit time duration on a 4-Mbps Token Ring network and the value 1/16,000,000 to BITTIME(2) to represent the bit time duration on a 16-Mbps network.

Referring to our previous nine-step approach used in the development of a 4-Mbps Token Ring performance model, step 2 computed the station delay. Using the variable S.DELAY to represent the station delay, we obtain:

In step 3 we determined the time to place a token on the ring. Using the variable T.PLACEMENT to represent the token placement time, we obtain:

Then, using the variable H.TRINGFLOW to represent the time for a token to be placed on the network and traverse half the ring (step 4 in our earlier model), we obtain:

Once a token is grabbed, it is converted into a frame. Because the time required to place the frame onto the ring depends on the length of the frame, let us use the array variable FRAMELENGTH(F) to denote different frame lengths. In actuality, let us assign different information field values to each FRAMELENGTH(F) value and add 21 bytes to represent the overhead per frame. Then, if we use the variable FRAMETIME to denote the time required to place a frame on the ring, we obtain:

As noted in step 6 in our prior model, the frame must traverse N/2 stations, on average, to reach its destination. If we denote the variable H.FRAMEFLOW to represent the time required for the frame to be placed on the ring and flow N/2 stations down the ring, we obtain:

If we use the variable C to denote the cable length (ring plus twice each lobe distance) in 1000- foot increments and the variable C.PROPTIME to denote the propagation delay time, we obtain:

Then, to compute the frame rate using the variable FPS, we obtain:

8.1.9 Program TPERFORM.BAS

To facilitate the execution of our general Token Ring performance model, the program TPERFORM.BAS was developed. This program, whose statements are listed in Table 8.1, can be used to generate a series of tables that indicates the frame rate based upon the network operating rate, number of stations on the network, average frame length, and network cable length in 1000-foot increments. As previously noted, the frame length is specified in terms of the information field to which 21 bytes representing frame overhead are added.

Table 8.1: Program Listing of TPERFORM.BAS
 REM PROGRAM TPERFORM.BAS         CLS REM THIS PROGRAM GENERATES A SERIES OF TABLES INDICATING THE FRAME REM RATE ON A TOKEN-RING NETWORK BASED UPON THE NETWORK OPERATING REM RATE, NUMBER OF STATIONS, AVERAGE FRAME LENGTH AND TOTAL NETWORK REM CABLE LENGTH      FOR K = 1 TO 7                 ' initialize frame lenghts      READ FRAMELENGTH(K)      NEXT K      DATA 64,128,256,512,1024,2048,4096      BITTIME(1) = 1/4000000      ' initialize bit duration      BITTIME(2) = 1/16000000      RATE$(1) = "4MBPS"            ' initialize network rate      RATE$(2) = "16MBPS" START:      LCOUNT = 0                    ' initialize line count      FOR I = 1 TO 2                ' vary network operating rate      IF I = 1 THEN GOTO NXT      FOR LC = 1 TO 50 - LCOUNT: LPRINT : NEXT LC: LCOUNT = 0 NXT: GOSUB HOUTPT                  ' print page header      FOR N = 10 TO 260 STEP 10     ' vary number of stations      FOR F = 1 TO 7 STEP 1         ' vary frame length (bytes)      FOR C = 2 TO 10 STEP 2 ' vary cable length (per 1000 feet)      S.DELAY = 2.5 * BITTIME(I)      T.PLACEMENT = 24 * BITTIME(I)      H.TRINGFLOW = (N/2) * S.DELAY + T.PLACEMENT      FRAMETIME = (FRAMELENGTH(F) + 21) * 8 * BITTIME(I)      H.FRAMEFLOW = (N/2) * S.DELAY + FRAMETIME      TOTAL.TIME = H.TRINGFLOW + H.FRAMEFLOW + (N/2) * S.DELAY      C.PROPTIME = C *.00000164#      FPS = 1/(TOTAL.TIME + C.PROPTIME)      GOSUB DOUTPT      NEXT C      NEXT F      NEXT N      NEXT I      END HOUTPT:      LPRINT "FRAME RATE OF A "; RATE$(I); " TOKEN-RING NETWORK"      LPRINT "BASED UPON THE NETWORK OPERATING RATE, NUMBER OF"      LPRINT "STATIONS, FRAME LENGTH AND TOTAL CABLE LENGTH "      LPRINT      LPRINT "NUMBER OF  AVG FRAME   CABLE LENGTH FRAME RATE"      LPRINT "STATIONS    LENGTH       X000 FEET   IN FPS"      RETURN DOUTPT:      IF LCOUNT < 50 THEN GOTO SKIP      FOR LC = 1 TO 10          ' move to top of next page      LPRINT      NEXT LC      LCOUNT = 0      GOSUB HOUTPT SKIP: LPRINT USING "  ####    ######"; N; FRAMELENGTH(F);       LPRINT USING "          ###"; C;       LPRINT USING "      ######## "; FPS       LCOUNT = LCOUNT + 1       RETURN 

In examining the program listing of TPERFORM.BAS, note that the FOR-NEXT loops that vary the number of stations, frame length, and cable length would result in 910 frame rate computations for each ring operating rate. Rather than place a large set of tables in this chapter, the results obtained from the execution of TPERFORM.BAS were placed in the BASIC directory at the previously noted Web URL. In addition, readers can modify the entries in the DATA statement to initialize a specific frame length more applicable to their network, or change the FOR-NEXT loop variable values for N and/or C to obtain information concerning the frame rate for a specific number of stations or cable length that is not in the table.

To provide readers with an example of the ease with which TPERFORM.BAS can be modified as well as data we can use to further discuss Token Ring performance, let us modify the program. By changing the FOR N loop parameters to 40 TO 50 STEP 5 and the FOR 5 loop parameters to 1 TO 2 STEP 1, the printed outputs contained in Table 8.2 and Table 8.3 are obtained.

Table 8.2: Frame Rate of a 4-Mbps Token Ring Network

Number of Stations

Avg Frame Length

Cable Length 1000 Feet

Frame Rate (fps)

40

64

2

4613

40

64

4

4544

40

64

6

4477

40

64

8

4413

40

64

10

4350

40

128

2

2900

40

128

4

2873

40

128

6

2846

40

128

8

2820

40

128

10

2794

45

64

2

4515

45

64

4

4449

45

64

6

4385

45

64

8

4323

45

64

10

4263

45

128

2

2861

45

128

4

2835

45

128

6

2809

45

128

8

2783

45

128

10

2758

50

64

2

4422

50

64

4

4359

50

64

6

4297

50

64

8

4237

50

64

10

4179

50

128

2

2824

50

128

4

2798

50

128

6

2772

50

128

8

2747

50

128

10

2723

Frame rate of a 4-Mbps Token Ring network based on the network operating rate, number of stations, frame length, and total cable length.

Table 8.3: Frame Rate of a 16-Mbps Token Ring Network

Number of Stations

Avg Frame Length

Cable Length 1000 Feet

Frame Rate (fps)

40

64

2

17651

40

64

4

16685

40

64

6

15819

40

64

8

15039

40

64

10

14332

40

128

2

11280

40

128

4

10877

40

128

6

10503

40

128

8

10153

40

128

10

9826

45

64

2

17293

45

64

4

16365

45

64

6

15531

45

64

8

14778

45

64

10

14095

45

128

2

11133

45

128

4

10740

45

128

6

10375

45

128

8

10033

45

128

10

9714

50

64

2

16950

50

64

4

16057

50

64

6

15253

50

64

8

14527

50

64

10

13866

50

128

2

10989

50

128

4

10607

50

128

6

10250

50

128

8

9917

50

128

10

9604

Frame rate of a 16-Mbps Token Ring network based on the network operating rate, number of stations, frame length, and total cable length.

8.1.10 General Observations

In reviewing the results of the frame rate computations represented in Table 8.2, let us first examine the effect of a change in the average frame length versus a change in the cable length of a network. This will enable us to determine the relative effect of the average frame length versus cable distance for a network with a given number of stations.

For a 40-station network with an average frame length of 64 bytes, note that each increase in network cabling by 2000 feet results in a decrease in the frame rate ranging from 69 (4613 ˆ’ 4544) to 63 (4413 ˆ’ 4350) frames per second. Note that a 40-station network with an average frame length of 128 bytes has a decrease in the frame rate ranging from 27 (2900 ˆ’ 2873) to 26 (2820 ˆ’ 2794) frames per second as the cable length increases in 2000-foot increments from 2000 to 10,000 feet. When the average frame length is 64 bytes, a decrease in frame flow of 64 frames per second per 2000-foot cable length increase is equivalent to 64 * 64 * 8, or a decrease of 32,768 bits per second in the information flow capability of the network. When the average frame length is 128 bytes, a decrease in frame flow of 26 frames per second due to a cable length increase of 2000 feet results in a decrease of 128 * 26 * 8, or 26,628 bits per second in the flow of information. Thus, as the average frame length increases, the effect of an increase in the amount of cabling used in the network slightly decreases.

Now let us examine the frame rate as the average frame length increases and the number of stations remains fixed. Note that for a 40-station network, an increase in the average frame rate from 64 to 128 bytes for a cable length of 2000 feet results in a decrease in the frame rate of 1713 (4613 ˆ’ 2900) frames per second. For a cable length of 10,000 feet, an increase in the average frame length from 64 to 128 bytes for a 40-station network results in a decrease in the frame rate of 1556 (4350 ˆ’ 2794) frames per second. Thus, the effect of the frame length on the frame rate exceeds the effect of the cable length.

Now let us turn our attention to observing the effect of an increase in the number of network stations with a fixed average frame length and cable length. For a 40-station network that has an average frame length of 64 bytes and a cable length of 2000 feet, the frame rate is 4613 frames per second. When the number of stations is increased to 45, the frame rate drops to 4515, a decrease of almost 100 frames per second. On a per-station-increase basis, this results in a decrease of approximately 20 frames per second when the average frame length is 64 bytes. Note that this decrease in the frame rate is slightly less than the decrease in the frame rate as the network cable distance increases from 2000 to 10,000 feet. This means that each increase in the number of stations has a lesser effect on network performance than an increase of 8000 feet in the cabling used in a network. Although these figures slightly differ as the number of stations on a network increases, you can use the preceding as a general guide for configuring and expanding Token Ring networks. That is, by limiting your network cabling distance, you may be able to alleviate the effect of an increase in the number of network stations on network performance.

8.1.11 Station Effect on Network Performance

To obtain a more detailed understanding of the effect of an increase in the number of network stations on network performance, the comprehensive series of tables contained in the BASIC directory at the noted Web URL was used to extract data. In doing so, the frame rate for 4- and 16-Mbps Token Ring networks was extracted for 64-byte frame lengths and 10,000 feet of cable as the number of network stations varied from 10 to 260 in increments of ten stations. Table 8.4 contains the frame rate information extracted from the comprehensive table.

Table 8.4: Frame Rate versus Number of Network Stations

Frame Rate (fps)

Number of Stations

4 Mbps

16 Mbps

10

4956

15938

20

4736

15364

30

4535

14830

40

4350

14332

50

4179

13866

60

4022

13430

70

3876

13020

80

3740

12634

90

3613

12271

100

3495

11928

110

3384

11603

120

3280

11296

130

3182

11005

140

3090

10728

150

3003

10465

160

2921

10215

170

2843

9976

180

2769

9748

190

2699

9530

200

2632

9322

210

2569

9123

220

2508

8932

230

2451

8748

240

2396

8573

250

2343

8404

260

2293

8241

Based on a 64-byte average frame length and 10,000 feet of network cabling.

In examining the frame rates for 4- and 16-Mbps networks listed in Table 8.4, note that the number of stations has a considerable effect on the information flow on a Token Ring network. For example, a ten-station 4-Mbps network supports a frame rate of 4956 frames per second, which is equivalent to an information transfer rate of 4956 frames/second * 64 bytes/frame * 8 bits/byte, or 2.537 Mbps. For a 260-station network, the frame rate is reduced to 2293 frames per second, which is equivalent to an information transfer rate of 2293 frames/second * 64 bytes/frame * 8 bits/byte, or 1.174 Mbps. Turning our attention to the frame rates listed in Table 8.4 for a 16-Mbps Token Ring network, note that a ten-station network supports an information flow of 15,938 frames/second * 64 bytes/frame * 8 bits/byte, or 8.16 Mbps. When the number of stations is increased to 260, the information rate decreases to 8241 frames/second * 64 bytes/frame * 8 bits/byte, or 4.219 Mbps. Although the primary reason a Token Ring network supports a maximum of 260 network stations is based on jitter of the bits flowing on the network, the approximate halving of the information transfer capability is another important consideration for limiting the number of stations on a network.




Enhancing LAN Performance
Enhancing LAN Performance
ISBN: 0849319420
EAN: 2147483647
Year: 2003
Pages: 111
Authors: Gilbert Held

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net