2007

First Task and Process Management Serer for the Mexokernal ProX, Praveen, K(03C70), Subramani, R(03C95), Vanniaperumal, S(03C100)

ProX is a MeXokernal which is a sophisticated blend of microkernal and exokernal. The kernal is very thin and does efficient synchronous message passing and multitasking. The kernal with a robust and state of the are desig nneeds a complete innovatively designed process management server, Memory Management Server and Device driver tailor-made for the new MeXokernal. ProX is more like a microkernal but exposes the hardware to certain trusted tasks like thtat of the exokernal. The First task is the first user level task started by the kernal and it is in-charge of all the other management servers loaded as modules by the GRUB. Any process to be run is mapped through this mother of all processes by placing an aentry in the task table. The process management server creates , controls and schedules the processes through the two stage priority based scheduler. It is also responsible for the Process Table containing information on all processes in the system.

A Framework for Memory Management Server For Mexokernal ProX Harini.V, Lakshmi.N, Logeswari.P.V

This project is a framework of a memory management server suitable for an operating system that is built upon ProX, a mexokernal based system. Memory Management Server is implemented that runs as a privileged task in the user space.

Device Driver For Mexokernal ProX Muthusivam A (03C54), Senthil Kumaran S (03C86), Sriram R - April 2007 (03C91).

The main Objective of this project is to create a device driver framework for Goatee and to implement a sample device driver using the framework. Goatee is a 32 bit operating system for the Intel 386+ architecture. It is built upon mexokernel ProX. In this module the divice driver framework is created and a simple keyboard driver is implemented. It takes the advantage of the mexokernel design to the fullest.The device driver framework would essentially deal with the mechanism followed by the kernel to handle control to the device driver task, and how the device driver task interacts with the kernel after processing the interrupts. The main objective is to develop a simple, modular and clean code for the operating system.The complete development of the project was done in Debian GNU/Linux platform. The entire project is implemeted using C programminglanguage and Assembly. The entire source code of the implementation is available under the anonymous Subversion at the following location

Performance Effective Task Scheduling for Parallel Systems, Suguna, M, (06CS24)

A high performance computing environment is a suite of processors interconnected by high-speed networks, thereby promising high speed processing of computationally intensive applications with diverse computing needs. A well known strategy behind efficient execution of a huge application is to partition it into multiple independent tasks and schedule such tasks over a set of available processors. Such a partitioned application can be represented by a Directed Acyclic Graph (DAG). Finding an optimal scheduling for an application modeled by a Directed Acyclic graph (DAG) is known to be NP-Complete. This study deals with a simple scheduling algorithm based on list scheduling, namely, Performance Effective Task Scheduling (PETS) algorithm for homogeneous computing systems with complexity O((v + e)(p + log v)), which provides effective results for applications represented by DAGs.

Adaptive Fuzzy Neural Control For Nonliner System Senthil Kumar.S (06CS20)

The learning ability of Neural networks has been widely recognized as a powerful tool in Industrial control image processing applications etc. This work proposes a self organising adaptive fuzzy-neural control for aclass of non-linear systems. The proposed self organising adaptive fuzzy neural controller comprises of a computation controller comprises self organising fuzzy neural network identifier and it is the principal controller, which is used to estimate the controller dynamics with the structure and parameter learning phases. The structure learning phases possess the ability of generation and elimination of fuzzy rules. The supervisory controller is used to reduce the effects of approximation error and it guarantees the stability of the sytem by Lyapunov function. Structure determination and approximation ability are the major concern of my project. Since fault detection is a nonlinear system problem, this concept is used to detect rotor air gap eccentricity fault occuring in an Induction motor.