Converged networks: The future?
Here are some global trends in the communication arena, which are leading to the evolution of converged networks and the challenges involved in building such networks

next generation networks must be capable of handling voice, video, and data communications with proper QoS in a packet-based transport/switching infrastructure

In the future, when data will become the dominant component, it will be more practical to carry voice over data networks

The formation of the Bell Company in 1876 marked the arrival of the Public Switched Telephone Network (PSTN). From a network that supported just voice calling, it has today evolved to offer advanced services. It represents $250 billion in network investment and hundreds of billions of dollars in annual revenues.

The Internet originated in the 1960s as ARPANET from a packet-switched research project sponsored by Advanced Research Project Agency (ARPA).

IP networks and PSTN employ conceptually different architectures.

The PSTN uses a switched circuit network architecture (SCN) that is an end-to-end connection from the calling party to the called party. This includes two distinct layers, the circuit layer and the switched transport layer. With all signaling traffic traveling on the Signaling System 7 (SS7) control network, transport resources are freed up from signaling traffic. PSTN consists of large, centralized, proprietary Class-5 switches with remote switching modules (RSMs) and digital loop carriers (DLCs). PSTN offer very high reliability, excellent QoS for voice communications and speech delivery with exceptional clarity.

On the other hand, IP networks employ connection-less, "best-effort" Packet Switched Network (PSN) architecture, which offer very little or no QoS guaranties. Though the PSDN user-base is substantially smaller than PSTN, it is growing at an exponential rate.

Telecom Reforms-The Catalyst
The U.S. Telecommunications Act of 1996 prompted global trends towards deregulation and liberalization of telecommunications. This Act is now widely expected to be the beginning of a new age of innovative and affordable telecom services.

But consumers are yet to see any of this happen in a substantial way. The primary reason for this failure is that though the policies have changed, there is little change in the underlying infrastructure of the service providers and new service creation continues to take a 12 to 18 month period.

Network providers at the transport layer see their product as 70 percent network with a 30 percent wrap of services, but customers view the same product in the reverse proportion.

Network providers dealt with this rise in data by carrying it on the analog and TDM networks designed and optimized for voice or by building separate, parallel networks for high capacity data traffic. In the past, when voice traffic dominated the network, this approach made sense. In the future, when data will become the dominant component, it will be more practical to carry voice over data networks the "convergence networks".

Convergence Networks The Interim solution
To protect multi-billion dollar investment in their existing networks, operators will prefer a gradual transition to an all-packet next generation network, spread over a decade or more, instead of whole scale replacement. Hence, some sort of "convergence network" (next-generation data services layered on top of the existing networks) to bridge the gap between PSTN and emerging PSDN is necessary.

The converged network will allow co-existence of PSDN, PSTN and also the calls that originate on the IP-side of the network to terminate on the PSTN-side and vice-versa. Voice over IP (VoIP) uses the Internet Protocol (IP) to transmit voice as a packet over an IP network. The voice signals are digitized, compressed and converted into IP packets before being sent across an IP network.

These new networks will be built on two basic building blocks a core transport/switching layer and an access edge. Initially, the core transport/switching infrastructure will be based on IP/ATM core switches, which form the backbone of the network with IP cores supporting Internet-related traffic and ATM cores carrying circuit-based traffic.

While IP and ATM will dominate the transport layer and are well suited for Class-4/toll circuit switching, Class-5 TDM will be required in the network for some time. So circuit switching and packet switching will coexist for some time, with IP, ATM, and TDM, all playing complementary roles. Eventually, IP and ATM networks will merge, with the ATM core becoming transparent and the IP becoming the principal protocol.

However, there are major challenges in building a reliable NGN that offers parity with PSTN in terms of QoS, ease of dialing, and convenience features such as call waiting, free phone. The first challenge is to have reliable call-signaling capabilities in packet switching. The second is to provision and control QoS. The third is to build a converged VoIP/PSTN as a transitional solution. The final challenge is to migrate to NGN that employs a comprehensive Packet Mode Architecture (PMA).

SoftSwitches The Heart of NGN
Traditionally, switch vendors used to supply proprietary applications that ran on Class 4 and Class 5 hardware. This meant long waits for the service providers who wanted new applications to differentiate their services from the competition. But this is all going to change with the Open Application Programming Interfaces (APIs) supported by Softswitch. These open APIs will allow service providers and third-party developers to develop innovative applications at their own pace without depending on switch vendors. Besides quick service creation, other benefits offered by Softswitch include service differentiation, interoperability across heterogeneous networks, and the smooth transition from circuit to converged networks and then to next generation of packet networks.

The Softswitch is software that resides on fault-tolerant servers in a highly scalable, reliable and distributed architecture environment and performs call control functions such as protocol conversion, authorization, accounting and administration operations. Because of its call control functions, the Softswitch must be able to communicate with both PSTN and IP networks which means Softswitch will have to interpret several protocols such as MGCP, H.323 and SIP.

Also creating a strong developer community for Softswitch API is an equally important factor in its eventual success.

In spite of these issues, Softswitch is destined to become the most crucial component of next generation networks and is expected to give telecommunications the same rate of growth the Internet experienced in the 1990s.

QoS A Challenge
Quality of Service (QoS) is important for the converging networks to be successful. Besides simple assurance, QoS should be clearly defined, quantified, easily expressed and packaged simply enough for people to work with, and pay for.

QoS is about providing an environment that preserves the isochronicity of a signal i.e. the signal goes in and comes out, undistorted, perceptually in "real-time." In terms of IP voice and real-time media, this means developing strategies to reduce and regulate end-to-end delays such as codec and packetization delay, transmission and buffering delay, and routing delay. Also of concern are packet loss and jitter.

QoS encompasses all seven layers of the Open Systems Interconnection (OSI) Model, as well as every network element from end-to-end, operating systems (real-time scheduling, threads), communications protocols, data networks, scheduling and traffic management issues.

The concept of M3 multi-service, multi-technology and multi-vendor is driving and enabling Next Generation Networks. But providing QoS assurances in this mixed environment is exceedingly difficult. Each service type has a different concept of what constitutes high quality. Voice services can accept signal loss to some extent, but very little time delay or jitter in order to ensure high call quality. On the other hand, data services can accept time delays to some extent, but not packet loss.

Though QoS and bandwidth are related, increasing bandwidth to achieve QoS is not going to solve the problem since it is likely that new applications will fill the available bandwidth in no time. So even with fiber-to-curb and purely optical networks, certain issues may never be resolved.

The fundamental trade-off in quality of service is between bandwidth and traffic engineering and prioritization. As bandwidth becomes more constrained, traffic engineering and prioritization become more important. This trade-off in turn leads to two global QoS strategies one linked to bandwidth (RSVP), and the other prioritization (Diffserver).

Resource reservation (e.g. RSVP) is where network resources are allocated according to an application's QoS request, subject to bandwidth management policy. Prioritization (e.g. DiffServ) is where network traffic is classified and network resources are allocated according to bandwidth management policy. In the latter case, each application traffic flow has an associated Class of Service (CoS), based on which the network elements give preferential treatment to the classes having more stringent requirements.

The type of QoS that is more suitable for an individual flow depends on the kind of application and the network topology. The most popular QoS protocols are ReSerVation Protocol (RSVP), Differentiated Services (DiffServ) and Multi Protocol Label Switching (MPLS).

ReSerVation Protocol (RSVP)
RSVP, which has been used for videoconferencing for years, provides the signaling to enable network resource reservation generally on a per-flow basis. RSVP emulates the closest thing to PSTN's circuit switching on IP networks. The router must maintain the state information for each flow to allocate resources based on available capacity. RSVP lacks scalability and good policy control mechanisms.

Differentiated Services (DiffServ)
DiffServ provides a coarse and simple way to categorize and prioritize network traffic. Each packet is marked to belong to a particular class of service and is sent through the network. Each router in the path inspects the packet header and determines "packet scheduling". This eliminates the need for maintaining individual flows on all routers. DiffServ system supports "admission control" (refusing customers when the capacity is not available), "packet scheduling", "traffic classification", and "policies" and "rules".

Multi Protocol Label Switching (MPLS)
MPLS provides bandwidth management via network routing control, according to labels inside (encapsulating) packet headers. MPLS routing establishes "fixed bandwidth pipes" similar to ATM virtual circuits, but with "coarser levels" of QoS compared to ATM.

MPLS is now becoming a de facto standard and may emerge as "the winner" in the QoS protocol race. However, despite these differences, MPLS and DiffServ are not mutually exclusive.

Conclusion
Eventually the networks of the next generation must be capable of handling voice, video, and data communications with proper QoS in a packet-based transport/switching infrastructure. It must provide parity with PSTN in terms of QoS, ease of usage, and convenience features such as call waiting.

The next generation networks' revolution is not about the convergence of voice and data, but about packet-based broadband networks. Unlike, the B-ISDN, information superhighway, and Internet II, NGN is not going to fade away primarily due to the maturity of the underlying technologies, cost savings offered by migrating to packet network infrastructure and revenue opportunities provided by explosive growth of data in the core network.

V. V. Nageswara Rao,Engineering Manager, Huawei Technologies, prd@in.huawei

>>