Maximizing Primary Capacities in Survivable Networks

Maximizing Primary Capacities in Survivable Networks

Arun K. Somani (Iowa State University, USA) and David W. Lastine (Iowa State University, USA)
DOI: 10.4018/978-1-61350-426-0.ch007
OnDemand PDF Download:
No Current Special Offers


Achieving low blocking probability and connection restorability in the presence of a link failure is a major goal of network designers. Typically fault tolerant schemes try to maintain low blocking probability by maximizing the amount of primary capacity in the network. In this chapter, we assume the total capacity on each link is fixed, and then it is allocated into primary or backup capacity. The distribution of primary capacity affects blocking probability for dynamic traffic. This can be seen by simulating dynamic traffic with different ways to distribute capacities in a network. A Hamiltonian p-cycle is a capacity optimal way of allocating primary and backup capacity. However, different Hamiltonian p-cycle may deliver different blocking probability for dynamic traffic. In general, more evenly distributing the backup and primary capacity lowers the blocking probability. This chapter provides upper bounds on how much primary capacity a network can provide if it uses a link based protection strategy to guarantee survivability for one or more link failures. Using integer linear programs we show that requiring preconfiguring carries a cost in terms of capacity if the solution is structured as a set of cycles.
Chapter Preview

Introduction: Survivability And Capacity Planning

The problem of providing dependable connections in networks is receiving more attention due to growing transport bandwidths and the consequent huge losses associated with failures of components in such high speed networks. While the current deployed networks operate at 10 Gbps speeds, networks operating at 40 Gbps are also gaining in popularity. The high vulnerability of the networks can be seen in that failures have been observed to happen as often as once every four days (Frederick, Datta, & Somani, 2006). Several researchers have studied dependability problem in various context.

Utilizing Primary Capacity Effectively

Today the usable bandwidth even in a single fiber is more than a user requires. To maximize the primary capacity usable in a network, the bandwidth must be shared. Several techniques exist to allow bandwidth sharing. Many of these techniques can be used simultaneously.

One way to share bandwidth is to use different wavelengths for different connections. Two connections using different wavelengths can use the same fiber at the same time as long as the frequencies are sufficiently different that receiver nodes can distinguish between them. Depending on the minimum spacing between distinguishable frequencies this method is referred by names such as wavelength division multiplexing (WDM) or dense wavelength division multiplexing (DWDM). Past a certain intensity of the electric field, the response of the fiber becomes non-linear. This results in phenomena such as four wave mixing, which prevent some combinations of wavelengths from being usable. In this chapter we assume wavelengths have been spaced appropriately to avoid this.

Optical transmitters and receivers components may be designed for a fixed frequency or be tunable over a range of frequencies. Signals in the optical domain can be routed using an Optical Cross Connect (OCX). Complexity of hardware to route optical signals varies in the degree of granularity for routing wavelengths. Some hardware can route individual wavelengths while other hardware routes groups of wavelengths referred to as a waveband. Support for splitting a signal to create a multicast tree exists. Optical or electronic amplifiers maybe required to maintain signal viability.

A second way bandwidth can be split on a fiber is by the different users of a fiber taking turns transmitting. This technique is known as time-division multiplexing (TDM). Of course TDM and WDM can be done at the same time, in which case it's the users of a frequency that take turns using the frequency.

Reserving frequencies and time slots for point to point connections can result in bandwidth being wasted when a network has bursty traffic since one connection may have extra capacity at one instant while another connection needs more capacity. For bursty traffic, use of a light trail can maximize the amount of capacity available to serve traffic. A light trail is a unidirectional optical bus that allows for all nodes on the bus to timeshare the bus. Since a light trail has multiple possible source nodes, it has a greater chance of using available capacity than a dedicated light trail between two nodes.

Another way of sharing bandwidth is to use code division multiplexing. This method allows multiple users to use a wavelength at the same time by using different codes. Codes spread out data bits into many chirps. For some or all codes, receivers can detect a specific code by looking at the correlation of chirps over time. Increasing the number of users that can share a wavelength results in codes which need more time to transmit.

Complete Chapter List

Search this Book: