Most enterprises had recently lost their stamina to manage and control an optimized integrated voice and data network. Instead they had opted to handle voice via virtual services such as SDN, VPN or VNet and focused solely on their data intranets to handle their growing data traffic. This phenomena had made the networking gurus to predict the demise of all integrated voice/data networks. They had already forgotten the economies of scales associated with T1 backbones that allowed TDM-based integration of voice and data over backbone trunks only.

However, changing environments and newer technologies are bringing about challenging opportunity to those who have the patience and guts to study newer alternatives to meet the user needs in a cost effective manner. Nowadays, multimedia sources are a rule rather than an exception. At each site, voice, data, image and video traffic sources generate traffic for the enterprise intranets, extranets and of course the internet. Now it is possible to integrate voice and data over subscriber lines (SLs), access lines (ALs) and trunks (TKs) using either IP or ATM technologies and save a significant amount of money. Whereas in the olden days of T1 networking it was possible to get data traffic to ride free over an existing voice network, it is now possible to get voice traffic to ride free over an existing enterprise intranet. In other words, the days of packetized voice have finally arrived.

It should be interesting to study the feasibility of integrating voice over an existing data intranet maintained by a small enterprise consisting of 17 sites with a headquarter located at Las Colinas, Texas. The busy hour voice traffic intensity (TI) in MilliErlangs and other pertinent network design data are shown in the database Slide 1. Before studying the benefits of integrating voice and data, we should first design an optimum voice network topology by modeling the use of voice grade (VG), 56Kbps and T1 lines. The software package EcoNets was employed to model nine (9) topologies using optimally located 1, 2 and 3 switches. The computations for Total Cost/Month and Cost/CM are shown in the next two Slides 2 and 3. The analysis shows that an optimal network topology consists of two switches, one at West LosAngeles California and one a Las Colinas and uses 56Kbps ALs and TKs. The design employs voice-digitization/compression similar to ITU's G.729 (about 7000 bps per voice conversation ) and circuit-switching technologies. In other words, a 56Kbps circuit can carry 8 voice conversations. The optimum voice network topology is illustrated in the next Slide 4.

The analysis shows that one can achieve six cents per call minute on a standalone voice network handling about 625800 call minutes per month. This value is already comparable (if not cheaper than) to those currently available from major long distance carriers. Such a standalone voice network has sufficient excess capacity to handle the data traffic as defined by the database of Slide 1 with BHrTI now measured in bps. One can test this hypothesis by adding the total traffic (in bps) due to voice and data applications and comparing it with the capacity of ALs and TKs of the optimized voice network (Slide 4). By creating a new VHD input file with total traffic ( in bps) and using it to create a IVD network topology one will also discover that the monthly cost of this network remains the same as that for an optimized voice network. It should be emphasized that voice network still employs circuit switching and that data rides on the star topology. EcoNets can be employed to show that a standalone optimum data network will cost about $10,000 per month or about $11,300 per month based on multi-drop (MD) and star topologies respectively.

We can follow the above procedure to study the performance of several data intranets based on four traffic intensity levels, with and without the digitized voice traffic. We will employ four models of traffic intensities (TIs). Consider the data traffic multiplying factors (TMF) of 30, 100, 200 and 300 to obtain new data TI values from the DB Slide-1 where MErl values are converted to bits/sec for this study. The new traffic represents multimedia traffic to and from the corporate internet/database server located at Las Colinas. The digitized voice traffic (in bps) considered in the previous circuit switched study is now doubled to consider the effects of overheads required by IP/UDP or ATM protocols. It should be pointed out that a data TMF of 30 represents a case where the data and voice traffic are close to being equal.

Several data networks based on the 2-switch star topology were modeled with and without digitzed voice traffic. In order to obtain the least cost solution for each traffic volume, five link types (56Kbps, 112KbpsFT1, 225KbpsFT1, 512KbpsFT1 and T1) were considered for both ALs and TKs. The best case solutions are shown in the next Slide 5. It shows that the cost of a call minute varies between 0 and 2.2 cents depending upon traffic multiplication factor (TMF) and link types used. (It should be borne in mind that a non-optimum network topology can result in a cost of 8 cents per call minute). The packet delays vary in range of 25-50 ms when T1 circuits are employed. This should ensure good quality of service for both voice and data.

In conclusion (Slide 6), the integration of packetized voice with enterprise data intranets is an excellent idea. The age of packetized voice has finally arrived and circuit switching in no longer a viable solution for a typical voice application. It should be stressed that the above networking models are valid for enterprise intranets based on either IP (also called VoIP) or ATM technology (also called Voice/Telephony Over ATM or simply VTOA). This topological study shows no bias towards any technology as long as it is cost effective and the user is satisfied.
 

For additiohnal information on this study, please email to: rsharma@econets.com