Networking with Bypass and Multiple T1 Data Links.
Almost any T1 network with more than two nodes can save on the mileage charges for lines by designing the topology of the network to minimize the number of lines and their lengths. This usually means a series of shorter segments connecting nodes in sequence, for example Chicago--Dallas--Atlanta--New York (Figure 1).
Connecting between the Nodes
But what if Node A (Chicago) wants to talk to Node C (Atlanta), or a node several segments away? One way would be to set up a direct circuit between each pair of nodes, leading to a mesh design, which can be quite expensive. There's no need to run a separate line from A to D if the multiplexers at B and C can somehow take data coming in on one T1 line and send it out on another. This form of networking can be done inside a single multiplexer or between two multiplexers, back to back.
Simple muxes, even those limited to a single data link, can pass on data if the channel is first demultiplexed so it appears at a local port, for example at B (Figure 2). This channel can then be patched into a port on a second mux at B, (Figure 2). This channel can then be patched into a port on a second mux at B, the one connected to the single data link running to C. Naturally, this technique requires, at B, two multiplexers, two ports (one in each mux) and a cable for each channel passed from A to C.
The smarter muxes pass channel between data links without demultiplexing. That is, input to a port at Node A can output on a port at Node C without appearing on a local port at Node B. What this process is called depends on your historical roots. The details of how Node B works vary significantly, depending on the type of multiplexer.
Supporting Multiple Data Links
Slightly more-sophisticated multiplexers support multiple data links, but act in other respects the same as point-to-point muxes. That is, data channels must be terminated in a port at every node. To go from A to C, data multiplexed at A must be sent over the first line, demultiplexed at B, brought out on one port, cabled into a second port on the same piece of equipment, and multiplexed again for the trip to C (Figure 3).
Even some relatively new T1 multiplexers use this technique, particularly where the T1 multiplexer is configured internally as faster and slower multiplexers in tandem. A group of channels from one T1 composite, destined for the other T1 circuit, is brought out at a port on the link-rate multiplexer. Instead of being split further by the subrate multiplexer, this port is cabled into another high-speed mux port (Figure 3). Such a "group bypass" might carry 64 to 512 kb/s of digital bandwidth between T1 circuits.
Coming from the Phone World
If one comes from the telephone world, where T1 has long been used to handle digitized voice, the process is "drop-and-insert." The T1 line at Node B is viewed as coming in from Node A and going out to Node C. Even though the two branches to A and to C appear as separate four-wire circuits at Node B, and attach to the multiplexer by separate connectors, they're considered to be only one line (Figure 4).
A D3 or D4 channel bank is the specialized multiplexer that always deals with voice channels digitized at 64,000 b/s each. Wne inserted in this "single line" at Node B, some advanced channel banks have the ability to drop off a conversation from A at B, and insert another conversation from B to C. Channels originating at A that are not specifically interrupted at B pass on to C without being demultiplexed. Internally, these drop-and-insert channel banks are usually configured as twin multiplexers with group bypass.
Channel banks are limited--to two T1 connections, to 64-kb/s channels, and in other ways.
From a data communications viewpoint, this A-B-C arrangement consist of two T1 lines: A-B and B-C. A data mux in this configuration is said to have multiple data links. The T1 multiplexers with multiple data links, described above as "more sophisticated," perform drop-and-insert on all channels; that is, every port at A is dropped at Node B into a local port, and likewise for ports linked to Node C. With this third type of mux, connections from A to C are madw ith a patch cable between two ports at B, working with demultiplexed data (Figure 3).
Likening to a Voice Channel Bank
This is essentially the same drop-and-insert found in a voice channel bank. In effect, there are two point-to-point multiplexers in the same box. the ideal would be to move data in one link and out another without needing local ports or patch cables. In the data world, this feature is known as "bypass," a term coined less than 10 years ago to distinguish what was then the novel and flexible handling of data from the rigid specification for digitized voice.
The most-modern T1 facilities-management systems perform true bypass. They avoid the demux-patch/remux process by transferring the data for specified channels directly between T1 lines. This process takes place on a logic card, and so does not require a pair of local ports or a patch cord. One card handles a large number of channels (Figure 5).
A voice multiplexer assumes that most conversations are just passing through, and should come in one T1 line and go out another. It must do something (drop-and-insert) to terminate a channel locally.
Assuming All Terminate Locally
A data multiplexer assumes that all channels should terminate locally. It does something (bypass) to route a channel in one line and out another.
Conceptually, drop-and-insert see a multiplexer as sitting in the middle of a continuous line. Bypass positions the mux between adjacent lines, which are segments in a larger network.
Mechanically and electrically, the line connections are the same. From a certain viewpoint, then, one might be tempted to believe that drop-and-insert can be treated and used interchangeably with bypass. This would be a mistake, as will be developed below.
In the preceding discussion, the definition of a channel has been assumed. In most cases, this also is a mistake. Again, it depends on whether the discussion grows from a voice or a data background.
Converting Analog to Digital
In the voice world, conversations always start as analog signals derived from the spoken word. But in transmission, many of them are converted to digital signals (processes covered by a separate technical backgrounder, available from me on request).
The world standard for digitized voice is pulse code modulation (PCM) at 64,000 bits per second. Most often, the conversion from analog to digital (and back) is done by a specialized multiplexer, the D4 channel bank. A channel bank always works iwth PCM, so there are "always" 24 standard voice channels in one T1 line (Figure 6).
It's important for voice transmission on the public switched telephone network that every channel be the same, that is, 64,000 b/s. Switch equipment in the central office (CO) must deal with a huge number of calls; and for maximum speed and efficiency, every call must be the same.
Working with standard voice channels, the switch and DACS (digital access and cross-connect system) in the CO can route calls without demultiplexing them. In effect, the switch and DACS drop, insert and bypass on a grand scale. That's the reason for such a rigid standard (DS-1) for voice channels multiplexed on a T1 carrier.
DS-1, for "digital signal level 1," defines the format for sending 24 voice channels digitized at 64,000 b/s each, plus telephone signaling information (such as ringing and busy), synchronization patterns and various diagnostic information over a T1 circuit.
Finding a "Fact" Flawed
The wide use of the DS-1 standard has established in most users' minds the "fact" that there are 24 channels in a T1 link. This needs to be true only if those channels carry information intended to be routed over the public switched telephone network (PSTN). (As a digital signal, the information could be voice, data, facsimile, or anything.)
The standard does not apply fully on lines leased from common carriers, not at all in privately owned networks, nor anywhere else that the digital information is not intended for the PSTN. The portions of the standard regarding framing do apply to T1 lines supplied by AT&T Communictions after January 1, 1985, but not necessarily the specification for 24 channel must be 64,000 b/s.
The viewpoint of a channel from a data background is quite different. For data, any end-to-end connection is a channel, not matter what its speed. A data channel can run at standard rates from 50 b/s on up to, and beyond, the T1 rate of 1.544 megabits per second.
Calculating the Data Channels
Then how many data channels are there on a T1 line? It depends.
By straight time-division multiplexing (each channel on a line given a fixed, not necessarily equal, portion of the bandwidth), there could be 160 channels at 9600 b/s on one T1 circuit. This is a very real possibility, as all T1 multiplexers are TDMs, though few of them support as many as 160 channels. If each of those 9600-b/s channels were further subdivided by statistical multiplexers or TDMs in tandem, the number of separate connections between devices (channels) could exceed 7500 (Figure 7).
Where did the 24 channels go?
In older-technology T1 multiplexers, where the equipment is based on a channel-bank design, there remains the requirement that the bandwidth be cut first into 24 pieces. Most old designs impose restrictions on how each 64,000-b/s channel may be further divided.
For example, in the worst case, a 64,000-b/s channel might be limited to one or two data channels, regardless of their speed. This T1 mux could then carry only 24 or 48 channels at 9600 b/s each (using a total of only 230.4 or 460.8 kb/s out of the 1,544 kb/s available). (Refer to Figure 8).
One specific product, when running its maximum 48 channels, is limited to 19.2 kb/s on each, or 921.6 kb/s total--less than 60 percent of the T1 bandwidth. The rest of the T1 circuit capacity is not available to the user.
Making Subchannels Available
Some older muxes can do better than that, making available several subchannels per 64,000-b/s channel, but usually not more than 56,000 b/s maximum. There's still that sharp boundary between channels, and it may waste some bandwidth even if the number of subchannels is not restricted.
Say a network runs 9600-b/s data channels exclusively. Dividing 64,000 by 9600, six of them fit, but there's a remainder of more than 6,000 b/s--that amount of bandwidth is not usable. The waste is repeated 24 times for a total of about 150,000 b/s, or 10 percent of the T1 capacity.
When there can be more than six subchannels at lower speeds in this type of multiplexer, the percent of bandwidth wasted varies, depending on the speeds of the subchannels (Figure 8).
Coming from a DAta Environment
The modern T1 facilities-management system comes from a data environemnt. When it was designed originally, there was no possibility of asking a central office to switch data channels on the public telephone network (though this seems to be coming). T1 facilities-management systems were aimed at private networks, where the user performs any switching needed.
The result is that data muxes could be made with efficiency and flexibility to get the maximum throughput over a T1 line. The division by 24, a requirement for voice on the public switched network, was simply irrelevant.
Network designers can build extraordinary networks at relatively low cost with T1 multiplexers having multiple data links and the ability to bypass data.
Explaining How the Costs Drop
First, the costs come down.
The bandwidth of the T1 is often less expensive, per bit, than multiple analog lines or DDS (Dataphone Digital Service) circuits of the same aggregate capacity. Even if this were not the case, T1 might still be used for increased reliability or convenience in network management.
With multiple data links, each site or node requires only one piece of equipment. It's not necessary to buy a mux for each T1 line going to another location; so for each additional line after the first, there's a significant saving from needing one less mux.
Bypass substitutes a single logic card for any number of channels that formerly required pairs of ports with an interconnection cable for each. Not only does the user buy fewer expansion modules, the reduction in cabling means there's an improvement in reliability.
If many ports were to be bypassed, the savings in ports and expansion modules would reduce the requirements for rack space, number the requirement for rack space, number of mux chassis, and more, for sizable reductions in the overall cost of the network.
Second, the design becomes considerably more flexible.
Spelling Out the Flexibility
In dropping and inserting a 9600-b/s data channel, a modern mux (such as the Timeplex Link/1) requires bandwidth on the link of only 9600 b/s plus some overhead, not 64,000 or some fraction of the voice-channel standard. This means that the full capacity of the liens is utilized very efficiently, and the network designer need not worry about the boundaries of 24 channels (Figure 8).
Bypass allows on T1 data link to serve many locations, reducing the total number of lines required. This feature is important even if the network normally does not require bypass, because it makes possible alternate data paths around failed network elements, thus improving reliability.
The prime example is the simple triangular ring network. Each link carries data between a pair of nodes, not aware of the third node. But if one link fails between two nodes, there's an alternate path through the third node--if that third node can bypass the data channels (See Figure 9).
The Link/1 ssytem provides for automatic alternate routing. In the event the normal path is unavailable, due to failure of a line or a multiplexer, connections are established over those routes that are operating.
Routing via a Patch Panel
The switchover to alternate routes also uses another feature of the latest multiplexers: soft routing, or an electronic patch panel under the control of the operator at the central side. It's not necessary that port one at Node A connect to port one at Node B; any centra port can connect to any port at the remote end.
Several alternate configurations of the network can be planned out in advance and stored in memory. Any one of them may be invoked later by a single operator command, or automatically by a time-of-day clock.
By taking full advantage of the features offered by modern T1 multiplexers, data communications users can reduce their costs, reduce the amount of equipment needed, and increase overall reliaiblity with more tolerance to failed lines and nodes.
This article focused on data. But one of the biggest advantages of T1 networks is they can carry voice as well as data. The economics of T1 networking are combined.
But that's another story.
|Printer friendly Cite/link Email Feedback|
|Date:||Apr 1, 1985|
|Previous Article:||Vendors Are Urged to Provide Improved Customer Service.|
|Next Article:||All You Ever Wanted to Know Concerning System Acquisition.|