In distributed control, the control functions are shared by many processors within the exchange itself. This type of structure owes its existence to the low cost microprocessors. This structure offers better availability and reliability than the centralized SPC.

Exchange control functions may be decomposed either 'horizontally' or 'vertically' for distributed processing. In vertical decomposition, the exchange environment is divided into several blocks and each block is. Assigned to a processor that performs all control functions related to that block of equipments. The total control system now consists of several control units coupled together. The processor in each block may be duplicated for redundancy purposes and operates in one of the three dual processor operating modes discussed in centralized SPC. This arrangement is modular so that the control units may be added to handle additional lines as the exchange is expanded.

In horizontal decomposition, each processor performs only one or some of the exchange control functions. A typical horizontal decomposition is along the lines of the functional groupings shown in Fig. a chain of different processors may be used to perform the event monitoring, call processing and O&M functions. The entire chain may be duplicated as illustrated in Fig for providing redundancy. Similar operating principles as in the case of dual processor structure apply to the dual chain con­figuration.

Level 3 Processing

Since the processors perform specific functions in distributed control, they can be specially designed to carry out these functions efficiently. Level 3 processor handles scanning, distribution and marking functions. The processor and the associated devices are located physically close to the switching network, junctors and signaling equipment. Processing operations involved are of simple, specialized and well-defined nature. Generally, pro­cessing at this level results in the setting or sensing of one or more binary conditions in flip-flops or registers. It may be necessary to sense and alter a set of binary conditions in a predefined sequence to accomplish a control function. Such simple operations are efficiently performed either by wired logic or micro programmed devices.

A control unit, designed as a collection of logic circuits using logic elements, electronic or otherwise, is called a 'hard-wired' control unit. A hard-wired unit can be exactly tailored to the job in-hand, both in terms of the function and the necessary processing capacity. But it lacks flexibility and cannot be easily adapted to new requirements. A micro programmed unit is more universal and can be put to many different uses by simply modifying the micro program and the associated data. With the same technology, the micro programmed units tend to be more expensive and slower than hard-wired units for an equivalent processing capacity. When the processing is complex, microprogramming implementation is easier. Table below summarizes the cha­racteristics of micro programmed and hard-wired control. With the advent of low cost microprocessors and VLSI programmable logic arrays and controllers, microprogramming is the favored choice for level 3 processing.

In microprogramming, the binary conditions required for control functions are altered through a control word which contains a bit pattern that activates the appropriate control signals. By storing a set of control words in a memory and reading them out one after another, control signals may be activated in the required sequence. Recognition of this fundamental aspect of control leads to two approaches to the design of control word in a micro programmed system. The control word may be designed to contain one bit per every conceivable control signal in the system. A control scheme organized in this fashion is known as horizontal control. Alternatively, all the control signals may be binary encoded and the control word may contain only the encoded pattern. In this case, the control is known as vertical control. Horizontal control is flexible and fast in the sense that as many control signals as required may be activated simultaneously. But it is expensive as the control word width may be too large to realize practically. In vertical control only one signal at a time is activated and the time penalty to activate a set of signals may be unacceptably large. In practice, a via media solution is adopted where a control word contains a group of encoded words that permit as many control signals to be activated simultaneously. Some of the recent designs use standard microprocessors for scanning and distribution functions instead of designing a microprogrammed unit. The microprocessor based design is somewhat slower than the microprogrammed unit, and the latter is likely to dominate until low cost custom ICs for these functions become available.

Level 2 Processing

The processors employed for call processing in level 2 have, in most cases, been specially developed for this purpose in the past. Level 2 processor is usually termed as switching processor. Early general purpose computers were ill suited to real time applications and were large in size and expensive. With the arrival of minicomputers and then microprocessors, a number of real time applications outside the field of telecommunications have sprung up. This, in turn, has led to the appearance of standard pro­cessors suitable for real time applications in the market. Nonetheless, the exchange manufacturers have continued to prefer house-developed switch­ing processors for some time in order to maintain full control over the products and to contain the costs. Of late, however, the trend is to employ commercially available standard microprocessors for the switching processor functions.

Switching processors are not fundamentally different from general purpose digital computers. There are, however, certain characteristics that are specific to switching processors, as in the case of processors employed in process control or other industrial real time applications. Processor instructions, for instance, are designed to allow data to be packed more tightly in memory without unduly increasing the access time. Single bit and half-byte manipulation instructions are used extensively in switching applications. Special instructions for task and event queue management, which would enable optimal run times for certain scheduler functions, are desirable. The architecture of switching processors is designed to ensure over 99.9% availability, fault tolerance and security of operation.

The traffic handling capacity of the control equipment is usually limited by the capacity of the switching processor. The load on the switching processor is measured by its occupancy t, estimated by the simple formula

t = a + bN

Where

a = fixed overhead depending upon the exchange capacity and configuration

b = average time to process one call N = number of calls per unit time

The occupancy t is expressed as a fraction of the unit time for which the processor is occupied. The parameter a depends to a large extent on the scanning workload which, in turn, depends usually on the number of subscriber lines, trunks and service circuits in the exchange. The parameter value may be estimated by knowing the total number of lines, the number of instructions required to scan one line, and the average execution time per instruction. The estimation of the value of the parameter b requires the definition of call mix, comprising incoming, outgoing, and local and transit calls. This is because the number of instructions required to process each type of call varies considerably. For example, the number of instructions required to process an incoming call where there is no need to retransmit the address digits is much less than the number required processing a transit call. The result of a call attempt such as call put through, called party busy or no answer also affects the number of instructions to be executed. The number of subscribers with DTMF and rotary dial telephones and the percentage of calls to grouped (PBX) lines are also important factors. Taking these factors into account, a call mix may be worked out and the mean processing time per call attempt calculated, by taking the weighted average of the processing times for various types of calls.

Usually, the switching processor is designed to handle a traffic load which is 40% higher than the nominal load. When this overload occurs, the processor may be loaded only to 95% of its capacity so that traffic fluctu­ations can be absorbed.

0.95=a+1.4*b*N

Where

N_N = nominal load in terms of calls per unit time

Or

N­_N = (0.95-a)/1.4*b

The average instruction execution time is dependent on the instruction mix as different instructions take different times. The best way to evaluate a switching processor is to prepare a benchmark comprising representative call mix and measure the actual processing time under this load.

Level 1 Processing

The level 1 control handles operations and maintenance (O&M) functions which involves the following steps:

• Administer the exchange hardware and software.

• Add, modify or delete information in translation tables.

• Change subscriber class of service.

• Put a new line or trunk into operation.

• Supervise operation of the exchange. Monitor traffic.

• Detect and locate faults and errors.

• Run diagnostic and test programs.

• Man-machine interaction.

The complex nature of the functions demands a large configuration for the level 1 computer involving large disk or tape storage. As a result, O&M processor in many cases is a standard general purpose computer, usually a mainframe. The complexity and volume of the software are also the highest when compared to level 2 and 3 processing. The O&M functions are less subject to real time constraints and have less need for concurrent processing. Hence, it is a common practice that a single O&M computer is shared among several exchanges located remotely as shown in Figure below. In such an arrange­ment, the exchanges contain only the level 2 and level 3 processing modules. Remote diagnosis and maintenance permit expert maintenance personnel to attend to several exchanges from one central location.

Many exchange designs use a single computer located physically at the exchange site, to perform both the O&M and call processing functions. Some designs may use two different processors for O&M and call processing, but may not resort to remote O&M. Instead, they may use one dedicated O&M processor for each exchange.

Also see the following ppt.