-

The cost of internet bandwidth in Kenya, just like in other developing countries, is significantly high compared to the developed countries. The challenge of this research is to look at ways of making the most optimal use of the available bandwidth and improve performance in terms of reduced latency and better download speeds.
In this dissertation, we investigate various techniques that can be used to optimize the utilization of the limited and expensive international bandwidth in developing countries. We look at the effects that content distribution networks (CDNs) and cache servers have o international bandwidth utilization as well as latency. A model of determining the point at which a content distribution network would be cost-effective is formulated. We also look at the variation in latency for content that is found within the CDNs as opposed to the origin servers.
We have explored on the effect of local peering at internet exchange points and developed a model to determine the financial break-even point at which it makes a commercial sense for an ISP to join the internet exchange point in Kenya. We analyze the suitability of peering and transit models of internet exchange points as it applies to local Internet Service Providers (ISPs). Peering and transit models for inter-ISP traffic exchange are studied and we show the comparative difference each has on latency in accessing local content.
Noting that spam e-mails and viruses consume a significant amount of bandwidth, we explore the effect of content filtering of that junk at the upstream Internet backbone providers. We develop a model that local ISPs can use to determine the point at which it is cost effective to invest i a spam filtering solution located at a data center within the upstream internet backbone provider.
We analyze trends in internet bandwidth utilization and note that many organizations have very low utilization at night - traffic analysis and prioritization can ensure that less critical and bandwidth-hungry applications can make good use of this "night bandwidth"
We argue that as framework that combines each of these techniques would result in a significant reduction on the costs incurred by Internet Service Providers to pay for international transit bandwidth, a cost that they pass directly to the local end users. We look at the metrics that can be used to determine the optimal point of introducing each of thee bandwidth saving techniques. Data collected from a local ISP has been used to validate this framework. This demonstrates that such a framework can indeed provide significant savings on international bandwidth costs and at the same time improve the user experience. Internet service providers implementing such a frame work would ultimately be able to pass the cost saving benefits to their clients.