All IP networks are a work in progress

By Michael Kennedy

The FCC's efforts to develop a broadband policy for the U.S. have stimulated a policy debate as to the best way to extend broadband to 100 percent of Americans. AT&T aggressively entered the debate in an FCC filing by linking achievement of 100 percent broadband to the retirement of the Public Switched Telephone Network (PSTN) which it characterized as a relic of a bygone era.  Though there is a steady migration from the PSTN to all IP networks this relic of a bygone era still maintains higher reliability, quality and serviceability advantages over all IP networks. IP networking's quality gap must be closed before the 100 percent goal can be achieved.

The term all IP networks includes network element such as IP RAN (Radio Access Networks), IP DSLAMs, IP/Carrier Ethernet Access and Aggregation Networks, IP Service Edge routers, IP/MPLS Core Routers and Packet over Optical Transport.  PSTN network elements include analog twisted-pair copper subscriber lines, Circuit Switches, Digital Cross Connect systems, and TDM/SONET transport systems. The high quality and reliability of the PSTN is a consequence of its maturity, its reliance on standardized hardware, relatively simple software, and its origins in regulated monopoly telephone companies where quality, reliability and serviceability were valued over service innovation and low cost. In contrast, IP has its origins in a highly competitive, innovative, and tumultuous IT market where innovation, customer responsiveness and price competition were valued above quality, reliability and serviceability.

Systems vendors have invested heavily to make their IP network elements "Carrier Class." They recognize that service providers have significantly different requirements than the enterprise customers who were the early adopters of IP technology. We found in a recent survey, however, that there are many gaps remaining between service provider quality requirements and current IP network elements' capabilities.

IP networks have more complex software than the PSTN and IP network elements and interwork more extensively than those of the PSTN. Consequently, IP's quality, reliability and serviceability gaps are caused by the constant introduction of new software into the network and the complexity of managing the network from an end-to-end service perspective. We looked at quality and reliability from four perspectives-reliability and availability requirements, network element features, network operations and management features, and end-to-end availability features.

Reliability and availability are compromised by software defects in new software releases and by procedural errors. New software releases, upgrades and patches cause the majority of network outages. Software has become so complex that it is never bug free. Standards simply require that no critical defects exist upon software release and that defects at lower severity levels not exceed industry standard limits. Root causes of poor software quality include the many recent systems vendor mergers that have hurt internal communications and caused cuts in testing due to internal cost pressures.  Reliability also suffers through conscious decisions by systems vendors and service providers to shorten testing cycles to improve time-to-market for new features. Service provider procedural errors also compromise reliability and availability. Major causes of procedural errors are process complexity, lagging configuration tools, poor authentication and security, and a lack of process discipline in configuration changes. Some level of procedural errors once again may be unavoidable since the need to rapidly launch new service features is often a higher priority than fully testing the software.

Modular software, patch management and rollback features are required network element features to assure IP network availability and reliability. Modular software isolates problems and simplifies testing. Modularization also enables hitless software upgrades. A series of small software patches versus a full release upgrade simplifies the testing process and thus improves reliability. Rollbacks to earlier software releases are essential in limiting release installations to maintenance windows.

Network operations and management features include fault correlation and isolation, and element management systems. The key objective not yet met for IP networks is that element management systems need to move strategically from an element to element model to an end-to-end service management model using an improved management architecture that integrates well with service providers' custom OSS.

Finally IP network routing and other end-to-end route protection capabilities are essential to achieving end-to-end network availability. However, existing protocols converge too slowly especially as compared to recovery times in the PSTN. I believe that movement toward packet optical transport and the OIF's GMPLS control plane mechanisms combined with existing routing protocols will close this gap.

The majority of U.S. aggregation and access network facilities continue to use analog lines, circuit switching, T1 and SONET technologies. While migration to an all IP network is underway, IP-based networking quality, reliability and availability gaps remain as barriers to retirement of the PSTN. The incentives or unintended disincentives created by policy makers will influence the rate at which these gaps are closed.

Michael Kennedy is a regular FierceTelecom columnist and is the co-founder and Managing Partner of Network Strategy Partners, LLC (NSP) -management consultants to the networking industry. He can be reached at [email protected]