Automated, dynamic traffic engineering in multi-service IP networks [invited talk]

Since the initial conversion of the ARPAnet from the NCPfamily of protocols to the initial TCP/IP-family of protocols in the early 1980s, and especially since the advent of tools (such as browsers) for easily accessing content in the early 1990s, the Internet has experienced (and continues to experience) meteoric growth along any dimension that one cares to measure. This growth has led to the creation of new business segments (e.g. Network Element Manufactures and Internet Service Providers), as well as to partitioning of the operators providing IP services (access/metro/core). Many of the core IP operators also provide long-distance telephony services. Since the majority of their sunk costs are concerned with laying and maintaining optical fibre (as well as the equipment to light up the fibre plant), these operators would like to carry both classes of traffic (data and voice) over a single core network. Additionally, the meteoric growth in number of users forces these carriers, and their NEM suppliers, to continue to look for ways to obtain additional capacity from the installed fibre and network element plants. The net result of these pressures is that core IP networks have become increasingly complex. In particular, traditionally form of management products and processes are becoming less effective in supporting the necessary service provisioning, operation, and restoration. I will describe some of the research that Agilent Laboratories is pursuing to enable automated, dynamic traffic engineering in multi-service IP networks