The first section in this chapter focused attention on the development of mathematical models to estimate the flow of data on 4- and 16-Mbps Token Ring networks. This section uses the previously obtained knowledge about Token Ring performance to examine the flow of data between networks when bridges are used. In doing so, we obtain answers to such questions as: What is the forwarding and filtering rate required to be supported by a Token Ring bridge or router prior to that device becoming a potential network bottleneck? What effect does the use of different types of WAN transmission facilities have on the data transfer capability of Token Ring remote bridges and routers? Because the use of WAN transmission facilities is only applicable to remote bridges and routers, let us first examine the required level of performance of a local bridge and then focus on the use of remote bridges and routers.
Figure 8.1 illustrates the use of a local bridge to connect two Token Ring networks. Here, each Token Ring network can operate at either 4 or 16 Mbps, resulting in three distinct mixtures of network operating speeds the bridge must support: 4 and 4 Mbps, 4 and 16 Mbps, and 16 and 16 Mbps. Thus, the level of performance required by a local bridge linking Token Ring networks will depend on the average frame length, total cable length, number of stations , and the operating rate of each network. Obtaining this information will provide the frame rate in frames per second that you can expect each network to carry. Then, the frame processing capability of the local bridge should equal the sum of the frame rate expected to be carried by each Token Ring network or the bridge will become a network bottleneck.
Because the average frame length carried on a network varies by time and the number of stations on a network expands and contracts in tandem with personnel changes and corporate policy, instead of determining one number, you may prefer to work with a range of values. As an alternative, you may wish to select a network situation in which a bridge meeting the frame rate requirements of a certain networking scenario can be expected to satisfy a large variance in future changes to each Token Ring network to be interconnected . For example, assume you have a 4-Mbps and a 16-Mbps Token Ring network you wish to locally connect and, through monitoring, determined that the average frame length on each network was 128 bytes. Assuming the 4-Mbps network has 50 stations and the 16-Mbps network has 80 stations and each network uses 6000 feet of cable, you would use the Token Ring table at the Web URL to determine that the expected frame rate on the 4-Mbps network is 2772 fps (frames per second) while the expected frame rate on the 16-Mbps network is 9561 fps. Then, the local bridge should have a frame filtering capability equal to the sum of each network expected frame rate, or 12,333 fps. However, what happens if either network changes?
If the number of stations and length of network cabling increases, the frame rate on the network decreases. Thus, an increase in either or both of those variables has no effect on a bridge acquired to support the frame rate of smaller networks. Similarly, an increase in the average frame length results in a decrease in the frame rate on each network. Because the most variable factor is the average frame length, you may wish to consider selecting a lower average frame length, which in effect increases the network frame rate. In doing so, you are adding a margin of safety to the processing requirements of a local Token Ring bridge. For example, lowering the average frame length to 64 bytes while keeping the number of stations and cable length of the network constant results in the 4-Mbps network being projected to have a frame rate of 4297 fps, while the 16-Mbps network has its frame rate projection increased to 13,776 fps. Then, the local bridge you would use to connect the two networks should have the capability to process 18,073 64-byte frames per second. Concerning the bridge forwarding rate, under a worst-case scenario, assume that each frame on each network for a short period of time is routed to the other network. Then, the forwarding rate of the local bridge should equal its filtering rate to ensure the bridge will not become a network bottleneck and cannot congest the flow of data between networks.
As previously noted, the performance requirement for a local bridge is simply the sum of the frame rate on each connected network. That aggregate number of frames per second provides you with the overall frame processing rate a local bridge must exceed prior to it becoming a potential network bottleneck. In turning our attention to remote bridges and routers used to connect two geographically dispersed Token Ring networks, we must consider the operating rate of the WAN transmission facility used to connect networks. Unless that transmission facility operates at the network operating rate, a condition that essentially only occurs when a local bridge is used, the transmission facility functions as a filter, which reduces the maximum potential forwarding rate of frames routed from one network to the other network.
There are three key variables that govern the frame rate that can be achieved on a wide area transmission facility used to interconnect remotely located Token Ring networks. Those variables include the frame lengths, WAN operating rate, and the overhead of the transmission protocol used to transport frames between networks. Because the protocol overhead can vary, based on the type of protocol used by a bridge or router manufacturer as well as by their use of data compression, we will first examine the maximum frame rate based on different frame lengths without considering the effect of the protocol or presence or absence of data compression.
To facilitate the computation of frame rates based on different average frame lengths, the BASIC language program TRWAN.BAS was developed. Table 8.5 lists the statements in that program, which computes the frame rate for average frame lengths of 64, 128, 256, 512, 1024, 2048, and 4096 bytes. In actuality, the program adds 21 bytes to each frame length to correspond to the overhead bytes that are wrapped around the information field. Thus, the average frame length actually represents the length of the information field. The WAN operating rates, which vary from 4800 bps to 1.536 Mbps, represent commonly available analog leased line operating rates (through 19,200 bps) and commonly available digital leased line operating rates. Table 8.6 illustrates the table generated from the execution of that program.
REM PROGRAM TRWAN.BAS REM THIS PROGRAM COMPUTES THE MAXIMUM FRAME RATE THAT A WIDE AREA REM NETWORK TRANSMISSION FACILITY CAN CARRY BETWEEN TWO REMOTE BRIDGES CLS DIM MFPS(9, 7) 'max frames per sec FOR I = 1 TO 9 READ RATE(I) NEXT I DATA 4800,9600,19200,56000,64000,128000,256000,512000,1536000 FOR I = 1 TO 7 READ FLENGTH(I) NEXT I DATA 64,128,256,512,1024,2048,4096 FOR I = 1 TO 9 FOR J = 1 TO 7 MFPS(I, J) = (RATE(I)/8)/(FLENGTH(J) + 21) NEXT J NEXT I PRINT "WAN RATE --- --- --- --- MAXIMUM FRAME RATE PER SECOND --- --- --- --- --- -" PRINT " BPS BASED UPON AVERAGE FRAME LENGTH IN BYTES" PRINT " 64 128 256 512 1024 2048 4096 " PRINT J = 1 FOR I = 1 TO 9 PRINT USING " ######## "; RATE(I); PRINT USING " #####.## #####.## #####.##"; MFPS(I, 1); MFPS(I, 2); MFPS(I, 3); PRINT USING " #####.## #####.##"; MFPS(I, 4); MFPS(I, 5); PRINT USING " #####.## #####.##"; MFPS(I, 6); MFPS(I, 7) NEXT I END |
WAN RATE - - - - MAXIMUM FRAME RATE PER SECOND - - - - - - BPS BASED UPON AVERAGE FRAME LENGTH IN BYTES 64 128 256 512 1024 2048 4096 4800 7.06 4.03 2.17 1.13 0.57 0.29 0.15 9600 14.12 8.05 4.33 2.25 1.15 0.58 0.29 19200 28.24 16.11 8.66 4.50 2.30 1.16 0.58 56000 82.35 46.98 25.27 13.13 6.70 3.38 1.70 64000 94.12 53.69 28.88 15.01 7.66 3.87 1.94 128000 188.24 107.38 57.76 30.02 15.31 7.73 3.89 256000 376.47 214.77 115.52 60.04 30.62 15.47 7.77 512000 752.94 429.53 231.05 120.08 61.24 30.93 15.55 1536000 2258.82 1288.59 693.14 360.23 183.73 92.80 46.64 |
In examining the entries in Table 8.6, note that the T1 operating rate of 1.536 Mbps represents the T1 speed of 1.544 Mbps less the 8000 bps used for T1 framing. Also note that at the 1.536-Mbps rate and an average frame length of 64 bytes, the maximum frame rate of 2259 fps is less than the frame rate of 2293 fps for a 260-station, 4-Mbps Token Ring network constructed using 10,000 feet of cable. Thus, the WAN operating rate of a high-speed T1 circuit can be expected to function as a bottleneck for the transfer of data between Token Ring networks without considering the effect of the overhead of the transmission protocol or the use of data compression to enhance transmission efficiency.
To illustrate the effect of a protocol's overhead on the maximum frame rate supported by different WAN operating rates, let us revise the program TRWAN.BAS. If we assume the WAN protocol overhead is 20 percent, change the computation of the two-dimensional array MFPS to:
If you want to examine the effect of a different overhead percentage, change the multiplier 1.2 to 1.15 for computing a 15-percent protocol overhead, to 1.10 for a 10-percent protocol overhead, and so on. Table 8.7 lists the results obtained by executing the program TRWAN.BAS in which a protocol overhead of 20 percent was used for frame rate computations .
ADJUSTED FRAME RATE BASED UPON 20 PERCENT WAN PROTOCOL OVERHEAD WAN RATE - - - - MAXIMUM FRAME RATE PER SECOND - - - - - - BPS BASED UPON AVERAGE FRAME LENGTH IN BYTES 64 128 256 512 1024 2048 4096 4800 5.88 3.36 1.81 0.94 0.48 0.24 0.12 9600 11.76 6.71 3.61 1.88 0.96 0.48 0.24 19200 23.53 13.42 7.22 3.75 1.91 0.97 0.49 56000 68.63 39.15 21.06 10.94 5.58 2.82 1.42 64000 78.43 44.74 24.07 12.51 6.38 3.22 1.62 128000 156.86 89.49 48.13 25.02 12.76 6.44 3.24 256000 313.73 178.97 96.27 50.03 25.52 12.89 6.48 512000 627.45 357.94 192.54 100.06 51.04 25.78 12.95 1536000 1882.35 1073.83 577.62 300.19 153.11 77.33 38.86 |
Due to the ability of data compression to reduce redundancies contained within the information field of a Token Ring frame, the use of this technology increases the efficiency of the WAN transmission facility. Since 1990, several vendors have introduced compression performing remote bridges as well as added this capability to routers. Although the effect of compression depends on the level of data redundancy, which changes from frame to frame, most vendors advertise an average compression ratio of 2:1 for their products. What this means is that the logical data carried by the information field on the average is doubled within the physical information field. Thus, to consider the effect of data compression on the frame rate transfer capability of the WAN transmission facility, you would modify the computation of MFPS (I,J) in the program TRWAN.BAS as follows :
where K represents the reciprocal of the average data compression ratio. To consider both the effect of data compression and protocol overhead on the frame transfer rate of the WAN transmission facility, you would modify the computation of MFPS(I,J) as follows:
where K represents the reciprocal of the average data compression ratio and P represents the average protocol overhead.
To illustrate the use of the revised computation of MFPS(I,J), let us assume that the average data compression ratio is 2 and the expected protocol overhead is 15 percent. Then, K has a value of 1/2 or 0.5, while P has a value of 1.15. Table 8.8 illustrates the tabulation of frame rates obtained by modifying the computation of MFPS(I,J) as previously discussed and then re-executing the program TRWAN.BAS. In comparing the tabulations contained in Tables 8.4, 8.5, and 8.6, you will note that the key to reducing the potential of the WAN transmission facility from functioning as a network bottleneck is obtained by the application of data compression. In fact, at certain operational states, the WAN transmission facility may no longer be a bottleneck. For example, from Table 8.8 you will note that a T1 circuit is capable of transporting 3150 64-byte frames per second based on a data compression ratio of 2 and a protocol overhead of 15 percent. In comparison, we previously noted that the frame rate on a 260-station, 4-Mbps Token Ring network constructed using 10,000 feet of cable is 2293 frames per second. Thus, this WAN link connected to compression performing remote bridges would have the capability to transfer a greater number of frames per second than the 4-Mbps Token Ring network could generate. Readers can modify the program TRWAN.BAS and use it in conjunction with the entries in Table 8.8 to determine the most appropriate WAN operating rate based on the topology, size , and operating rates of networks to be connected as well as the transmission protocol overhead and the expected gain in transmission efficiency due to the use of compression performing bridges or routers, assuming this feature is included in the remote bridges or routers you are using or anticipate acquiring.
WAN RATE - - - - MAXIMUM FRAME RATE PER SECOND - - - - - - BPS BASED UPON AVERAGE FRAME LENGTH IN BYTES 64 128 256 512 1024 2048 4096 4800 9.84 6.14 3.50 1.88 0.98 0.50 0.25 9600 19.69 12.28 7.00 3.77 1.96 1.00 0.50 19200 39.38 24.55 14.01 7.53 3.92 2.00 1.01 56000 114.85 71.61 40.85 21.97 11.42 5.82 2.94 64000 131.26 81.84 46.69 25.11 13.05 6.66 3.36 128000 262.51 163.68 93.38 50.23 26.10 13.31 6.72 256000 525.02 327.37 186.75 100.46 52.21 26.63 13.45 512000 1050.04 654.73 373.50 200.91 104.41 53.26 26.90 1536000 3150.12 1964.19 1120.51 602.73 313.24 159.77 80.69 |
Now that mathematical models have been developed to estimate the frame flow on individual Token Ring networks and through the use of remote bridges and routers connected via the use of different types of WAN transmission facilities, let us focus attention on the traffic flow between remotely located interconnected networks. In doing so, we will use the model presented in Chapter 7, in which the total traffic flow, R T , between networks has the following relationship to the frame rates on networks 1 (R 1 ) and 2 (R 2 ) and the WAN transmission facility frame transfer rate, which we will denote as R 3 :
or
To illustrate the use of the preceding equation, let us assume you wish to download a 640-Kbyte file from a server located on a 16-Mbps Token Ring network that has 200 stations and 10,000 feet of cable to a 4-Mbps Token Ring network that has 50 stations and 4000 feet of cable, while the remote bridges or routers connecting each network use a 64-Kbps transmission facility. Because a file transfer uses the largest size frame available for most of the transfer, the transfer would take place using 4500-byte frames, which is the largest information field length supported by both 4- and 16-Mbps Token Ring networks. From the Token Ring table, the closest entries we can locate govern the frame rate for a frame length of 4096 bytes. From the table, the maximum frame rate supported by a 4-Mbps network with 50 stations and 4000 feet of cable is 121 fps, while the maximum frame rate on a 16-Mbps network with 200 stations and 10.000 feet of cable is 471 fps. From Table 8.7 we note that the maximum frame rate supported by a 56-Kbps digital circuit with a 20-percent protocol overhead is 1.42 fps when the average frame length is 4500 bytes. Using our formula to find the transfer rate, we obtain:
Similar to our notation in Chapter 7, the previously computed value of R T represents a best-case scenario in that it assumes the transfer of data from one network to another occurs without other stations on each network using the network or requesting inter-network transmission. Suppose the use of a monitoring program on each network indicated that at any time an average of three workstations were using the transmission facilities of each network and two workstations were using inter-network transmission. Then, you would adjust the values of R 1 and R 2 by dividing each by 3 and adjust the value of R 3 by dividing its value by 2. Doing so, we obtain:
Based on the preceding, you can expect a best-case frame flow between networks of 1.399 frames per second and an average frame flow between networks of 0.69 frames per second.
Based on an information field of 4096 bytes, the maximum flow of information between networks is 4096 bytes/frame * 1.399 frames/second, or 5730 bytes per second. Then, the optimum transfer time of a 640-Kbyte file would become:
To compute the average file transfer time, we would use the average transfer rate of 0.69 frames per second. Using an information field of 4096 bytes and a frame rate of 0.69 frames per second, the average information transfer capability between networks becomes 4096 bytes/frame * 0.69 frames/second. Then, the average transfer time to move a 640-Kbyte file between networks becomes:
This indicates that the transmission time to move a 640-Kbyte file between two Token Ring networks connected by a pair of remote bridges or routers using a 56-Kbps WAN transmission facility can be expected, on average, to require 232 seconds, or slightly under four minutes. In a best-case scenario, the transfer time will require 114 seconds, or slightly under two minutes. Both timings are based on the network characteristics previously discussed and will obviously change if either network or the WAN transmission rate changes.
You can extend the previous computations to determine a peak transmission time by monitoring each network to determine the peak number of users performing network activity at a given period of time. Once you obtain this information, you would then adjust the values for R 1 , R 2 , and R 3 similar to the manner in which those values were adjusted to determine the average frame transfer rate between networks. By considering the peak traffic on each network, your computation for R T would result in a worst-case frame transfer rate that, when used to compute the file transfer time, would provide the maximum transfer time.
Whether or not the optimum, average, and peak transfer times are acceptable depends on your organization and "its need for speed." While an increase in the WAN operating rate will have a significant effect on the ability to transfer frames between networks, it is not without cost. Thus, you may wish to perform a cost analysis to determine if a reduction in file transfer time or an increase in the transfer capability of the WAN is worth the expense associated with obtaining a higher WAN transmission capability. In addition, you should consider the relationship of file transfer activity in the form of programs and data files to the transfer of relatively short electronic messages in the form of files. For example, if the primary use of the Internet is to transport short e-mail files, would an additional expense of $1000 or more per month to reduce the transfer time of a message from three seconds to one second be worthwhile? Considering the fact that it may take a second or two for a person to move his or her hand to the keyboard and press a key to open the mail, I would probably defer the expense. However, the answer to such questions are for you to decide. In doing so, you can use the models provided in this book to determine transfer rates and transfer times, and then use that information in conjunction with the cost of communications equipment and transmission facilities as your decision criteria.