Cost Optimization

Among other considerations, one should not determine the placement of
network elements without considering the transmission requirements and
the likely transmission cost. For example in Figure 10-1, one could equally have determined that it would be better to place an MSC (and perhaps
some SGSNs) at the remote market in addition to just RNCs. This would
greatly reduce the transport requirements between the two cities. On the
other hand, however, there would be greater capital cost involved in placing
an MSC in the remote city. Alternatively, one could have decided that it
would be better to completely serve the remote city from equipment housed
in the larger local city. This would likely reduce the total RNC cost and
would avoid the need for a suitably conditioned building in the remote city
to house RNC equipment. The capital cost reduction in such a situation
could be considerable. On the other hand, the additional transport required
between the remote city and the MSC site could be very great and could
mean a large cost. (After all, there will likely be at least a T1 from every site
to the serving RNC regardless of how heavily used that site happens to be.
On the other hand, the Iu-Cs and Iu-PS interfaces are sized based upon utilization
only.)

Having said that, there will need to be a certain amount of transport
from the MSC to the PSTN in the remote market in any case. It may well
be that the size of that transmission facility is such that extra capacity is
available “for free” or that additional capacity can be added at a reasonably
low cost. Thus, the cost structure for transmission bandwidth must also be
considered. For example, although a DS3 supports 28 DS1s, the cost of a
DS3 is approximately 8 to 10 times that of a DS1. Thus, if one needs 12
DS1s, one is better off to lease a DS3 and get up to 20 DS1s “for free.” Similarly,
an OC-3 costs less than 2 DS3s, even though it supports up to 3 DS3s.
Finally, one must consider future technology evolution and the expected
costs and capacities of future network elements. If one were not anticipating
an upgrade to 3GPP Release 4, then the capital cost of an MSC in the
remote city might be justified if it could be depreciated over a seven- or tenyear
period and the effective cost compare with the transmission cost of
placing just BSCs or RNCs in the market. Imagine, however, that one is
deploying 3GPP Release 1999 and expecting to upgrade the network to
3GPP Release 4 within a two-year timeframe. In that case, one could delay
the deployment of switching equipment in the remote city until such time
as media gateways are available, provided of course that those media gateways
are sufficiently scalable and sufficiently inexpensive compared to a
traditional MSC. It might make more financial sense to absorb the cost of
transmission between the two cities until the more efficient architecture is
available.

Similar issues need to be considered in the placement of other network
nodes such as SGSNs or PDSNs, GGSNs, and so on. Let us take a UMTS example. An SGSN is at the same level as an MSC in the network hierarchy.
Consequently, it generally makes sense for SGSNs and MSCs to be colocated.
What about the placement of GGSNs? Well, that question comes
down to the types of data services that the network operator wishes to offer
and therelative use of those services. If for example a great deal of user traffic
goes to and from the Internet, then it would make sense to place GGSNs
at or close to the SGSNs and connect to the Internet relatively close to the
user. That can save bandwidth. On the other hand, if one expects that subscribers
will make a lot of use of operator-provided services, such as e-mail,
where those services are housed in a limited number of centralized locations,
then it can make sense to place the GGSNs nearer to those centralized
locations. Although that approach can mean greater transmission
overhead (because of the tunneling overhead between SGSN and GGSN), it
may also mean a net fewer number of GGSNs in the network. Given that a
GGSN or cluster of GGSNs needs to have other associated equipment, such
as DHCP servers and firewalls, a reduction in the number of GGSNs or the
number of GGSN locations may mean a considerable reduction in capital
cost. Again, we are faced with the issue of striking a balance between capital
expense and operating expense.

In the case of the placement of data nodes, there may also be special
cases that need to be considered. Imagine for example that a given network
operator establishes a relationship with a large corporate customer in a
given city. The individual subscribers from that customer may have a
totally different usage profile from other subscribers. They might, for example,
use the wireless data service exclusively for access to the corporate network.
In such a case it could be appropriate to dedicate one or more GGSNs
in a specific location for the use of those subscribers.