17.3 Connecting the cells to the switch
17.3.4 Protocols used for the interconnection
Clearly the base station and the switch need to use the same proto- col along whatever medium is used to interconnect them. The protocol Rolling Out the Network 269
1. To be more precise, those links at either end of the ring must be capable of carrying the total ring capacity; the one in the middle, half the link capacity; and those in between need a capacity increasing linearly between the center and the ends. That can be easily seen if it is imagined that either one of the links nearest the switch fails.
describes how speech and data calls are carried on the line and sends signaling such as establish call and disconnect call. If the switch and the base stations are purchased from the same manufacturer, the protocol may be irrelevant—it can be assumed that the manufacturer will provide equipment that will interconnect. The key issue is that some manufactur- ers have protocols that are efficient at compressing, or concentrating, the signals to be transmitted, reducing the bandwidth requirements on the link to the switch. Clearly, the more the signal can be concentrated, the more the cost can be reduced, particularly if leased lines are being used.
All those protocols run over E1 data links which are 2-Mbps (2.048 Mbps to be precise) links. The link is segmented into thirty-two 64-Kbps slots. Slots 0 and 16 are reserved for control data and synchro- nization information; hence, each E1 bearer can provide up to thirty 64-Kbps channels.
The following main protocols are available:
■ Channel-associated signaling(CAS);
■ Signaling system 7(SS7);
■ Q.931;
■ V5.1;
■ V5.2.
CAS can carry only POTS (simple voice) traffic, while Q.931 can carry only ISDN (data) traffic. CAS suffers from one key disadvantage: it is what is termed nonconcentrating. That means that for every subscriber, a time slot on the E1 bearer must be permanently reserved, regardless of whether a call for that subscriber currently is in progress. Because most subscribers generate traffic only around 5% of the time, that is extremely inefficient. Its only benefit is that blocking is not possible on the base station-to-switch interface.
The V5.1 protocol also is nonconcentrating, but it has the advantage of being able to carry both POTS and ISDN traffic. That is a significant advantage, as the following example demonstrates. In a network where both POTS and ISDN are provided (as will be the case for many WLL networks), if CAS and Q.931 are used, then separate E1 lines would be
required for CAS signaling and Q.931 signaling. If each required only half the capacity of an E1 line, then without V5.1, two E1 lines would be required, whereas with V5.1, only a single line would be required.
The V5.2 protocol is concentrating, so it is not necessary to dedicate resources on the E1 link to any particular subscriber—they can be assigned dynamically (in the same fashion as is performed on the air interface). This allows far fewer E1 links to be used, reducing the overall network cost.
At the beginning of 1997, the V5.2 standard appeared to be becoming widely accepted as the international means of connecting switches. Most manufacturers of WLL networks were in the process of developing a V5.2 interface for their products, due for release around the beginning of 1998.
However, there was also significant concern that there were many options in the way that a V5.2 interface could be configured and that V5.2 interfaces from different manufacturers would not necessarily interwork.
That matter needs to be monitored closely over the coming year, but at this point, V5.2 would appear to represent the best solution to interconnection.
If V5.2 is not used, it is possible that the system capacity is limited not by the air-interface capacity, but by the number of subscribers that can be accommodated on the E1 links to the switch. For example, in one system, the base station allowed a maximum of 16 E1 connections back to the switch using a V5.1 protocol. Given that each connection could support 30 subscribers, the total cell capacity was limited to 480 subscrib- ers. However, in a fully loaded system where the subscriber traffic is assumed to be low (0.1E) and the blocking tolerated high (1%), the air interface capacity could be as high as 850 subscribers. Therefore, the main point of blocking will be the switch connections, not the air interface.
In such a situation, some strange decisions are required. For example, rather than users being given a 2B+D ISDN connection, requiring two (or more) slots on the E1 interface, they would be given a B +D connection requiring only a single slot. They could transmit the same total amount of data, but it would take twice as long. The air interface traffic would be unchanged (just less bursty), while the E1 resourcing would be eased. Clearly, the designer of a network does not want to be placed in such a position.
Rolling Out the Network 271