PBX Call Processing Design Topologies

There is no PBX design standard that dictates the topology of the call processing network. Every PBX system has a common control complex that includes a Main System Processor, Main System Memory, and System Control and I/O Interfaces, but that may be the only common design element when comparing any two PBX system models. PBX call processing design topologies can be categorized into three general categories:

1. Centralized control

2. Dispersed control

3. Distributed control

These design topologies are shown in Figure 1.

Figure 1: PBX processing design topologies.

A fourth design topology can be added to this list—adjunct server control—if PBX systems equipped with an adjunct CTI applications server are taken into consideration, but only for support of features and functions beyond fundamental call set-up and connection functions. A CTI application server is totally dependent on the PBX common control complex to execute any and all communications operations. Adjunct server control design will be discussed separately later in this chapter.

Centralized Control

A PBX system with a centralized control call processing design topology includes the following processor elements:

1. Common control complex, including the Main System Processor

2. Port circuit card microcontrollers

In a centralized control design, the Main System Processor is responsible for all basic call processing functions and the control and execution of all switch network functions. This design specifically excludes local processors at the port equipment and/or port carrier shelf level. Port circuit card microcontrollers interface directly with the Main System Processor via a Service Control Interface.

A PBX based on a centralized control design may have additional processing elements to perform functions and operations not necessary to execute basic call processing functions, such as call set-up and switch connections, although they are necessary to ongoing call processing activities and system availability and survivability. The additional processing elements are typically used for administration, maintenance, diagnostics, and/or measurement operations. The Main System Processor usually handles some, or all, of these functions in small PBX systems with limited processing elements.

Centralized control designs are used most often by small PBX systems targeted primarily at customers with port requirements fewer than 200 stations, although there are a few notable exceptions. For example, all the Avaya Definity models are based on a centralized control design, including the very large G3r model, which can be equipped with more than 20,000 stations. PBXs based on a centralized control design must be equipped with a Main System Processor capable of handling all call processing and switch network functions, without diminished performance when the system is at maximum port capacity and all ports are active.

A centralized control design offers advantages and disadvantages. The major advantage of a centralized control design is fewer processor elements that can experience problems and disrupt service. Reduced local processor failures can increase overall system reliability and survivability. The major disadvantage is that the Main System Processor has total responsibility for all call processing and switch network functions, with no local processors to offset the processing load. Unless the Main System Processor is powerful enough to handle current and future call processing requirements, including support of new features and applications, there will be limitations on system performance levels. One solution to offload the call processing burden from the Main System Processor in a centralized control design is to install an adjunct CTI applications server in support of advanced feature and function needs that require significant processing power. There is a more detailed discussion of adjunct server control options later in this chapter.

Most PBX systems are configured at less than 50 percent port capacity, and the number of simultaneously active ports is almost always half that of the equipped ports. A Main System Processor should have little difficulty supporting the call processing requirements of 50 or 100 active ports, even if the processing element is a 16-bit or 32-bit microprocessor. Main System Processor problems can occur in large line size systems, with hundreds or thousands of active ports, if the main CPU is not equipped and designed to handle the potential traffic. When the Definity designers chose a centralized control design for the large and very large system models, a major design change from the older System 85, they also spent much time selecting the right processing element for the Main System Processor. The innovative selection of a Reduced Instruction Set Computing (RISC) microprocessor, the MIPS 3000, came at a time when all other PBX systems were based on a Complex Instruction Set Computing (CISC) microprocessor platform, such as the Intel 386 or Motorola 68030. A few years before Intel made its Pentium microprocessor commercially available, the MIPS 3000 was evaluated as the best available microprocessor in a centralized control design because it could handle the potential processing load at maximum equipped port capacities. The centralized control design resulted in call processing limitations for the Definity G3 models when configured for large, complex ACD-based call center installations. Definity is one of the few PBX systems that does not use an adjunct applications server to handle advanced ACD call analysis and routing functions, and the heavy processing load on the Main System Processor limits the system’s optional Expert Agent Selection (EAS) skill assignment programming parame- ters when compared with PBX systems equipped with an adjunct server to offload the processing burden.

Dispersed Control

A PBX system with a dispersed control call processing design topology includes the following processor elements:

1. Common control complex, including the Main System Processor

2. Local control processors at the port equipment cabinet and/or shelf level

3. Port circuit card microcontrollers

In a dispersed control design the Main System Processor is responsible for all basic call processing functions but may not execute all call processing and switch network functions. Local controllers provide the interface link between the Main System Processor and port circuit card microcontrollers and perform some call processing and switch network functions under the supervision and monitoring of the Main System Processor. The local processor elements function as slave controllers under the Main System Processor, which functions as the master controller. Like a PBX system based on a centralized control design, dispersed control designs can include additional processing elements typically used for administration, maintenance, diagnostics, and/or measurement operations.

A dispersed control design is the most common call processing design for intermediate to very large PBX system models. The primary advantage of a dispersed control processing design is that it offloads processing activities from the Main System Processor to increase overall system call processing performance. Call processing capacity is less dependent on the Main System Processor in a dispersed control design than in a centralized control design. The primary disadvantage is that failure or errors at the local processing level can affect service for all ports under its control. Unless the local controller is available in a redundant or duplicated mode, it is a potential major single point of failure for dozens or hundreds of system ports. For example, the Nortel Networks Meridian 1 Option 81C Controller Card can support one or two port carrier shelves, typically equipped with several hundreds of station ports. If the Controller Card, responsible for some call processing and switch network functions, fails, each port will lose service. The Controller Card is not available in a duplicated mode and is a major point of failure in the Meridian 1 Option 81C system design. Competing PBX system models from Siemens (Hicom 300H Model 80), NEC (NEAX2400 IMG), and Fujitsu (F9600 XL), for example, generally or optionally duplicate the local controller card at the port equipment cabinet/carrier shelf level to reduce the probability of service loss.

Distributed Control

A PBX system with a distributed control call processing design topology includes the following processor elements:

1. Multiple common control complexes, including multiple Main System Processors

2. Port circuit card microcontrollers

A distributed control design is based on peer-to-peer Main System Processors. It is similar in operation to a centralized control design because each Main System Processor is responsible for all basic call processing functions and the control and execution of all switch network functions. There is a major difference, however, in that each Main System Processor has control over a limited number of system ports. Each Main System Processor has responsibility for one or more port equipment cabinets but not all installed port equipment cabinets in the PBX configuration. This design excludes local processors at the port equipment and/or port carrier shelf level because the Main System Processor functions as a local processor to the one or two port cabinets it controls. In a distributed design port circuit card, microcontrollers interface directly with the Main System Processor via a Service Control Interface, if no local controllers are included in the design.

There are several important distributed control design advantages:

1. Multiple common control complexes increase system reliability and survivability; each Main System Processor is responsible for a limited number of system ports.

2. System call processing capacity is a function of the number of installed Main System Processors; adding an additional Main System Processor will increase the total system call processing capacity.

3. Multiple common control complexes are ideally suited for system configurations with multiple equipment rooms (campus, multilocation) because locations remote from the main equipment room are not dependent on a centralized Main System Processor for call processing operations.

A distributed control design requires synchronization and coordination among the multiple common control complexes for call processing, switching, and administrative functions. A true distributed control design supports transparent features and function operations across the system, with the option of using a single administration and maintenance interface for the entire system. The design is a difficult one to develop and operate successfully. One of the first distributed control designs was attempted by Rolm in the early 1980s, and the technical problems in synchronizing the VLCBX’s multiple Main System Processors and memory databases significantly delayed commercial availability of the product after its announcement. Rolm eventually solved the problems and was successful in marketing and selling its multinode CBX II 9000 system, a successor to the VLCBX, later in the decade.

Another early distributed control design that has been very successful and continues to be marketed and installed today is the Ericsson MD-110. First introduced in the early 1980s, the MD-110 was originally based on the Ericsson AXE 5 central office switching system design. Each LIM port equipment cabinet has a dedicated control complex; LIM cabinets communicate with each other over PCM-based FeatureLinks. In theory, a single MD-110 can be installed with more than 200 LIM cabinets and Main System Processors. LIM cabinets can be centralized or dispersed across multiple customer locations, with each cabinet dependent only on its local Main System Processor for all call processing functions. The MD-110 is currently the only circuit-switched PBX system based on a fully distributed Main System Processor design.

Legacy PBX Traffic Engineering Analysis

The call processing system of a PBX is responsible for all control functions and operations. It is responsible for monitoring and supervising all system design elements, including the switching network, printed circuit boards, and peripheral equipment, including station terminals and trunk circuits. It is also responsible for basic call set-up and teardown, feature/function provisioning, and systems management and maintenance operations. The PBX system call processing design has evolved from a single processing element responsible for all control functions to many processing elements, each with very specific responsibilities and functions. The PBX call processing design may differ from system to system, but there is a similar set of functional elements and responsibilities that is common to all enterprise voice communications systems, regardless of system size or functional complexity. Figure 1 illustrates the main PBX processing elements in a typical configuration design.

Figure 1: Main PBX processing elements.

Common Control Complex

The PBX common control complex can best be described as the brain of the system. Although other systems and subsystems may be responsible for physical operations, such as switch connections and voice signal transport, the common control complex is the command center responsible for issuing orders and supervising operations. There are several main components in a typical PBX common control complex, including:

• Main System Processor

• Main System Memory

• System Control Interfaces

• I/O Interfaces

The common control complex can be a single printed circuit board containing all of the listed common control elements, or it can be individual printed circuit boards for processor elements and interfaces and dedicated memory storage elements, such as hard disk or tape drives. The Main System Processor and Main System Memory are the two core elements in the common control. The System Control Interface and I/O Interface provide access to the two main common control elements for other internal system components and external system devices. The System Control Interface provides an intelligent link to the switch network (TDM bus) and processor bus to monitor port circuit activity and pass signals and messages between the main processor and local processors. The System Control Interface also can be used to support external call processing elements or adjunct systems dependent on the PBX common control complex. Examples are a CTI applications server used in call contact centers, and a third-party VMS. The I/O Interface ports typically are used for systems management, maintenance, and reporting functions. Examples are systems management terminals and call accounting systems. Figure 2 illustrates the main design elements of the common control complex.

Figure 2: PBX common controls and interfaces.

Defining PBX Traffic: CCS Rating

PBX traffic load is generally measured in 100 call-second units known as Centum Call Seconds (CCS). Centum from Latin, signifies 100. The maximum traffic load per station user during the Busy Hour is equal to 36 CCS, which is a shorthand method of stating 3600 seconds. Thirty-six CCS is equivalent to 60 minutes, or 1 hour of traffic load. A station port (telephone, facsimile terminal, modem, etc.) that “talks,” or connects, to the switch network for 10 minutes during 1 hour has a traffic rating of 6 CCS (10 minutes = 600 seconds = 600 call-second units). Combining the station user traffic load with an acceptable GoS level results in the following station user traffic requirement: 6 CCS, P(0.01). This notation signifies that a station user with an expected 6 CSS traffic load is willing to accept a 1 percent probability of call blocking when attempting to use the switch network. A 2 percent blocking probability would be expressed as 6 CCS P(0.02); a 0.1 percent blocking probability would be expressed as 6 CCS P(0.001).

A traffic rating of 36 CCS P(0.01) is used for station users who require virtually nonblocking switch network access. A 36 CCS traffic load is a worse-case situation because it is the maximum station user traffic load during the Busy Hour. The usual station user traffic rating requirement is about 6 to 9 CCS, P(0.01). Although a station user might be on a call that lasts for 1 hour or more—a 36 CCS traffic load—there is a very small probability all station users are simultaneously engaged in calls of at least 1 hour during the same Busy Hour. It is far more likely that an individual station user will have a 0 rather than a 36 CCS, traffic load during Busy Hour because that person may be in a meeting, traveling, on vacation, or too busy with paperwork to take or place telephone calls. Even if a station user makes several calls per hour, it is possible that each will be of short duration because many calls today are answered by a VMS with limited available time to leave a message. Most business-to-business calls today are connections between a station user and a VMS, and each of these calls typically last for less than 2 minutes and many last for less than 1 minute. An increasing number of callers no longer leave messages; they disconnect and send an e-mail.

The total PBX station traffic load during Busy Hour is simply the sum of the individual station user traffic requirements. If ten station users are connected to the network for 10 minutes during the same hour, the total traffic load on the switch network would be 60 CCS (10 station users × 10 minutes/station user, or 10 × 6 CCS). If the probability of blocking level was 1 percent, the traffic requirement would be noted as 60 CCS P(0.01). The total PBX station traffic load is rarely calculated, however, unless the switch network design is based on a single TDM bus or switch matrix. PBX traffic loads are better calculated for groups of station users sharing access to the same switch network element, assuming station users with similar traffic requirements are grouped together.

For switch network traffic engineering calculations, most customers use an average traffic load estimate to represent all station users instead of segmenting the station user population into like traffic load requirements. It is recommended that a different approach be used to traffic engineer a PBX system. Station users should be segmented into different traffic rating groups to ensure that switch network resources are optimized for each category of station user. In every PBX system there are some station users with very high traffic rating requirements, such as attendant console operators. Other station port types with very high traffic rating requirements include ACD call center agents, group answering positions, voice mail ports, and IVR ports. Each station port typically will have a 24 CCS traffic load, although customers usually prefer these ports to have nonblocking [36 CSS P(0.01)] switch network access and state so in their system requirements. Averaging the high traffic, moderate traffic, and low traffic station ports will result in a traffic engineered system that blocks an unacceptable percentage of calls for attendant positions because a rarely used telephone in the basement is using switch network resources instead of more important user stations.

As an example, a Nortel Networks Meridian 1 Option 81 C, based on a 120 talk slot Superloop local TDM bus design and a port carrier shelf that can typically support 384 ports, should be configured as follows to satisfactorily support the following station user traffic groups:

1. A maximum of 120 very high traffic station users, 36 CCS, P(0.01): stations configured on a single port carrier shelf supported by a dedicated Superloop bus

2. About 250 moderate traffic station users, 9 CCS, P(0.01): stations configured on a single port carrier shelf supported by a dedicated Superloop bus

3. About 500 low traffic station users, 4 CCS, P(0.01): configured across two port carrier shelves supported by a dedicated Superloop bus.

A single Superloop bus can adequately support each traffic group in this example, although the number of station users differs across the group categories. If the maximum number of potential traffic sources, or station users, is no larger than 120, then the Superloop bus is rated at 3,600 CCS, P(0.01). This is the maximum traffic handling capacity of a Superloop bus. The Superloop bus is rated at slightly less than 3,000 CCS, P(0.01), if the port carrier shelf is configured for about 256 station users, according to the original Meridian 1 documentation guide. If the number of potential traffic sources increases, then the traffic handling capacity decreases for a given probability of blocking level. The exact traffic rating for a specific number of station users is available with the use of a computer-based Meridian 1 configurator. Figure 1 illustrates CCS traffic handling capabilities of a Meridian 1 Superloop with 120 available talk slots. Customers with very high traffic requirements can configure a single Meridian 1 IPE shelf with up to four SuperLoops. Each SuperLoop is dedicated to four port card slots. Figure 2 illustrates how a port carrier shelf can be segmented.

Figure 1: Meridian 1 Superloop traffic handling capability.

Figure 2: Meridian 1 IPC module Superloop segmentation.

Traffic handling capacities for any PBX system local TDM bus are comparable in concept to Meridian 1 Superloop bus ratings:

1. If the number of potential traffic sources is smaller than or equal to the number of available talk slots, then station traffic can be rated at 36 CCS (nonblocking switch network access).

2. If the number of potential traffic sources is larger than the number of available talk slots, then the station traffic rating is less than 36 CCS. The traffic rating will decrease as the number of potential traffic sources increases.

The traffic handling capacity of the local TDM bus declines according to an exponential equation used to calculate probability of blocking levels. Most PBX designers assume a Poisson arrival pattern of calls, which approximates an exponential distribution of call types. The exponential distribution is based on the assumption that a few calls are very short in duration, many calls are a few minutes (1 or 2 minutes) in duration, and calls decrease exponentially as call duration increases, with a very small number of calls longer than 10 minutes. The actual traffic engineering equations (based on queuing models), call distribution arrival characteristics, and station user call attempt characteristics determining the local TDM bus traffic rating at maximum port capacity (if switch network access is not nonblocking) are known only to the PBX manufacturer.

Regardless of the actual traffic engineering equation used by the manufacturer, the calculated traffic rating will be based on three inputs:

1. Potential traffic sources

2. Available talk slots

3. Probability of blocking

A basic assumption used for most traffic analysis studies is a random (even) distribution of call arrivals during the Busy Hour. Traffic analysis studies also must make an assumption about call attempts that are blocked:

1. Station users who encounter an internal busy signal on their first call attempt continue making call attempts until they are successful.

2. Station users who encounter an internal busy signal on their first call attempt will not make other call attempts during a certain period.

In reality, station users who receive a busy signal will immediately redial. The assumption that a station user will not make another call attempt, if the first attempt is unsuccessful, is not realistic. The redial scenario is the assumption used by Poisson queuing model studies. The Poisson queuing model assumes that blocked calls are held in the system and that additional call attempts will be made until the caller is successful. For this reason, Poisson queuing model equations are commonly used by PBX traffic engineers to calculate internal switch network traffic handling capacities.

PBX systems with complex switch network designs (multiple local TDM buses, multitier Highway buses, center stage switch complexes) are far more difficult to analyze and traffic engineer than small PBX systems with a single local TDM bus design. Large, complex PBX switch network designs can provide a traffic engineer with many different switch connection scenarios that must be analyzed. Switch connections across local TDM buses require analysis of at least two switch network elements per traffic analysis calculation. To simplify traffic engineering studies, it is common system configuration design practice to minimize switch connections between different TDMs by analyzing call traffic patterns among stations users and providing station users access to trunk circuits on their local TDM buses. Centralizing trunk circuit connections may facilitate hardware maintenance and service, but it degrades system traffic handling capacity if more talk slots are used per trunk call.