Jyrki Akkanen

Case study of the evolution of routing algorithms in a network planning tool

Abstract Traffic routing is a key component in a network planning system. This paper concentrates on the routing algorithms and follows their evolution over multiple releases of a planning tool during a period of six years. The algorithms have grown from the initial stage of finding shortest paths with Dijkstras algorithm to cover more complex routing tasks such as finding protected and unprotected routes and capacity limited routing. We present the algorithms and in particular emphasize the practical aspects: how previous algorithms were reused and what were the practical experiences of using the algorithms. A conclusion of the study is that algorithms should be considered with an engineering attitude. It is not enough to focus on selecting the most sophisticated state-of-the-art algorithm for a given problem. Evolution capability, potential for reuse, and the development cost over the system lifetime are equally important aspects.

Reusing an open source application — practical experiences with a mobile CRM pilot

We discuss experiences in extending an open source CRM application to develop a new server-based mobile business application. Combining the application code reuse with incremental development process allowed successful development of a pilot application in a tight schedule. In particular, it enabled a quick start for customer-driven development, diminished risks related to the baseline application itself, and provided the flexibility needed in experimental pilot development.

Evolution of a software component - experiences with a network editor component

Even though the benefits of component-based software development are widely accepted, they are easily overestimated. To provide a firmer basis for the general discussion we describe our real life experiences with a software component. Having a lifetime of a whole decade the component has evolved from a class library to an independent component. In this paper we focus on the major evolution steps, their rational, and their outcomes, hoping that this gives some relevant insight to the issues that are important for software component evolution and maintenance. Surprisingly often the lessons learned have little to do with the hot topics of software technology that are being marketed. We discuss the risks attached to component selection, the usage of a shared platform for a product family, and the strengths and weaknesses of application frameworks and components. We also comment practical issues in designing and implementing major architectural changes.

Continuous Scheduling for Data-Driven Peer-to-Peer Streaming

The paper introduces a new approach to data-driven peer-to-peer media streaming. While classical algorithms send buffer maps and run scheduling periodically, we experiment by continuously sending incremental notifications of buffer contents and by running a continuous scheduling process. The approach has been tested both in a network simulator and in a small network of smart phones with live video content. The protocol seems to adapt quickly to changing data rate or churn without producing intolerable overhead. The approach leads to small latency and strong peer equality and may thus ease neighbor management. The end-to-end delay is small and one may alleviate a need for large buffers.