The Selection of the Proper Multiplexer Mandates a Careful Buying Evaluation.
Selecting a multiplexer from the low bidder is easy, but often costs more in operating efficiency than it saves in initial cost. Luckily, there is no need to understand the inner working of a multiplexer to make a more-careful selection. Efficiency and reliability depend on many factors, but they can usually be analyzed with a few straightforward questions. Adding flexibility and manageability to the decision process is not difficult and can lead to even greater operating efficiency.
Efficiency has always been the main reason for buying multiplexers. Telephone line costs are usually the largest expense in a data communication system, so efficient usage is of paramount importance. Unfortunately, the efficiency of today's statistical multiplexers is much harder to evaluate than it was for the older time-division multiplexers (TDM). Those TDMs allocated a fixed fraction of the phone line's bandwidth to each of the attached terminals.
Time-Division Comparison Is Easy
For example, a 9600-b/s line would be divided amoung four terminals at 2400/bs each. The best efficiency possible using this technology was 100 percent. In reality, efficiencies ranged from 95 percent to 99 percent because of the overhead information the multiplexer transmitted in addition to the data. Because time-division multiplexing is a relatively simple technique, efficiency was independent of how the multiplexer was used. Vendors published their efficiencies and comparison was easy.
In today's statistical multiplexer, or sta mux, there is no fixed allocation of the phone line's bandwidth. The stat mux takes advantage of the fact that asynchronous terminals are sending or receiving data only 10 to 25 percent of the time (synchronous terminals may be found to have higher utilization, but they still are well below 100 percent).
Stat-Mux Efficiencies Are High
By allocating bandwidth to a terminal only when it has data, that stat mux avoids wasting the expensive telephone line on idle time. This means that stat mux efficiencies can run up to 200 percent to 600 percent. For instance, 400-percent efficiency means that a stat mux with a 9600-b/s line can support 16 terminals at 2400 b/s. Since the terminals need bandwidth less than 25 percent of the time, the phone line is not overloaded.
A statistical multiplexer is far more efficient than a TDM, but comparing the efficiency of statistical multiplexers is difficult. Stat mux efficiency depends on both product design and on the data traffic being multiplexed. Any stat mux can be more efficient with terminals that pass data 10 percent of the time than with terminals that are used 25 percent of the time.
In addition, different multiplexers will show different efficiencies, even if they are handling identical data traffic. This is due to design differences. A design that's efficient for very low-usage interactive applications may not work as well in heavier-mode applications, and the opposite also may be true.
Terminal Use Makes a Difference
Since everyone uses their terminals differently, no definitive efficiency numbers can be published by multiplexer manufacturers. The efficiency specifications that are seen usually tell more about a vendor's assumptions of terminal utilization than about the performance of the multiplexer.
There are two ways to respond to this lack of information. One is to simply choose a product from a reputable vendor, and the other is to perform your own efficiency tests. The choice depends on how important it is to get the utmost phone-line efficiency.
A stat mux is so much more efficient than a TDM that most users will get all the efficiency they need simply by buying from an experienced multiplexer company. The leading manufacturers have refined their designs to provide good efficiency in a wide range of applications. Better efficiency is possible by choosing a product, such as the Codex 6002, that can be "tuned" for best performance. This tuning involves optimizing buffer allocation and the frame structure for a particular application.
For instance, an interactive terminal sends only a few characters at a time. This requires little buffer, but provides best performance when short frames are sent out frequently on the phone line. Blockmode terminals, on the other hand, send a whole screen of information at a time. This is best handled by providing the terminal with a large buffer and sending a few long frames over the telephone trunk. By tuning the multiplexer to your particular needs, you can avoid the performance compromises that occur when the same product is sold for a wide range of applications.
For situations where a more-thorough analysis is needed, the only reliable comparison is benchmark testing, using the actual data traffic to be multiplexed. This is expensive, because it involves getting equipment from several manufacturers, establishing a test facility, and using scarce personnel resources to conduct the tests. In practice, the vast majority of multiplexer users find that the efficiency they can gain by benchmark testing does not justify these expenses.
Purchase Decisions Can Be Simple
Despite the difficulty of evaluating the efficiency of modern statistical multiplexers, most purchase decisions can be made very simple. If data traffic is light, the product from any knowledgeable manufacturer will do an adequate job. In heavier traffic, it would be wise to look for a product that can be tuned to your specific usage pattern. Only in applications where even the last iota of efficiency is essential is it necessary to perform any of the benchmark tests.
Evaluating Dependability Isn't Easy
Data communications reliability is becoming more important, but evaluating dependability has not gotten any easier. Most multiplexer vendors use similar designs for similar products. Unless there is a clear difference in design, such as redundant circuitry, mean times between failure will be comparable. For most data communications professionals, it's more useful to focus on mean time to repair than on mean time between failure.
Each company provides a different level of service. Some swap-out defective units by mail, others send service technicians to the customer site to repair defective parts.
Mail swap-outs have a minimum wait of about 24 hours. If tha's acceptable, and if the equipment is easy to exchange, this can be an economical solution. If faster response is needed, or equipment is complex, on-site service is preferable.
However, it is essential that service personnel are near all locations in the data network, that the technicians are well trained, and that good spare inventories exist. Unless these conditions are met, on-site service may be no better than the mail. Since service costs can be high, it may prove worthwhile to keep your own spares for rapid repairs, and then have defective equipment repaired in the most-economical manner.
Service Performance Varies
Acceptable service performance varies with the importance of the applications running over a multiplexer. Often the best solution is to look for a company that offers a range of service options, so that you can have the most-economical solution that fits your needs.
A flexible multiplexer is one that keeps up with your evolving data communication needs. People insist on getting a multiplexer that works with their current data processing and terminal equipment. Buying a flexible multiplexer requires looking a little more closely, but can be far more economical. A simple change, such as adding a terminal, can be expected almost on a monthly basis. A flexible multiplexer can handle this extra terminal easily. Major changes, such as changing data processing equipment or adding new branch locations, happen less frequently--every year or two--but such changes will obsolete an inflexible multiplexer instantly.
The cost of inflexible equipment can take several forms. In its easiest form, it's simply the cost of the multiplexer that becomes obsolete prematurely. The more-insidious costs arise if the organization's data processing and communication structure is dictated by the multiplexer rather than by corporate needs. This can lead to losses in productivity that are far more costly than replacing a multiplexer. Delayed expansion, suboptimal data processing equipment selections and poor performance can result from not recognizing that the inflexible multiplexer is the real obstacle to change.
The importance of flexibility in a multiplexer is easy to see. Recognizing which multiplexer is most flexible is more difficult. Even though we all try to predict future changes, very few will base their decisions solely on their predictions. It's better to look for broad flexibility that will tolerate even unexpected changes. Specifically, this means looking for compatibility with a wide variety of data processing equipment, and the ability to operate in varied network configurations.
Even in the asynchronous environment, equipment compatibility is surprisingly complicated. Most multiplexers will handle the most-common asynchronous equipment. However, equipment with exceptional needs is becoming more common. Several vendors of word processing and graphics equipment have started to use an eight-bit-plus-parity character structure instead of the more-common ASCII seven bits plus parity. Other vendors are using proprietary data-restraint techniques instead of the common Xon/Xoff or other control-signal schemes. With the increased use of multiple minicomputers in a single organization, it is important to maintain the necessary flexibility to support all of this varied equipment within a single communications network.
Synchronous Is More Complicated
In synchronous environments, compatibility is even more complicated. True statistical multiplexing of synchronous protocols is available only from the more-sophisticated multiplexer vendors. Synchronous support in simple multiplexers is usually little more than time-division multiplexing. The increased use of synchronous protocols, especially in networking between minicomputers, makes it important to look for a multiplexer that can handle these protocols properly.
An expanding organization will outgrow even the most-flexible multiplexer eventually. Even stable organizations with rapidly growing computer usage will face this problem. A flexible multiplexer will not become obsolete, even under these circumstances. There are multiplexer manufacturers offering a full line of compatible products that allow today's multiplexer to be used as part of a larger network in the future. For even greater flexibility, look for a vendor with a flexible equipment-upgrade policy.
Management of a multiplexer network requires inceased time and skill as the network grows. Unfortunately, this becomes expensive. Hiring people with data communications expertise is expensive. Assigning this job to someone who already has a full workload can be even more expensive, because inadequate management can lead to performance and reliability problems. In choosing a multiplexer, these costs can be minimized by looking for work-saving features.
Central Control Is Essential
The ability to control an entire network from a single site is essential. The expense of having a data communications expert at both central and remote sites is unacceptable. Trying to control remote sites by telephoning instructions to inexperienced personnel at remote sites isn't much better. Most multiplexer vendors offer some central-control capability. Features to look for are the capability to perform exactly the same functions whether the operator is local or remote, and to have central control even if the net expands with other models by the same manufacturer.
A friendly operator interface is more than a convenience. A friendly interface allows fairly novice operators to take charge of the network, especially during off-peak hours. This assures that some network management will be available at any time it is needed, without burdening the organization with a large staff of expensive, hard-to-find data communications experts.
Data communications costs are constantly rising. Careful multiplexer selection can help keep these costs under control. Careful selection means that equipment should be viewed in many ways other than price. Luckily, it's possible to evaluate the critical areas of efficiency, reliability, flexibility and manageability with well-chosen questions, without having to understand the details of statistical multiplexer technology.
|Printer friendly Cite/link Email Feedback|
|Date:||Jan 1, 1986|
|Previous Article:||Micros and LSI Chips Are Fueling the Latest in Modem-Mux Advances.|
|Next Article:||Is This the Interconnect Service That Some Telecom Managers May Prefer?|