A novel multi-layer virtual memory system to solve memory leak problem

Other Title(s)

نظام مبتكر متعدد الطبقات للذاكرة الافتراضية لحل مشكلة تسرب الذاكرة

Dissertant

Atum, Ahmad Ali

Thesis advisor

Malkawi, Muhammad Isam

Comitee Members

al-Ajluni, Naim
al-Utaybi, Ghassan
al-Hajj Hassan, Muhammad

University

Amman Arab University

Faculty

Collage of Computer Sciences and Informatics

Department

Department of Computer Science

University Country

Jordan

Degree

Ph.D.

Degree Date

2007

English Abstract

Memory leak problem is one of the major causes of software failures.

Current approaches for solving memory leak problem are not thorough; they either detect memory leak in development environments using source code, re-linking, or recompilation or they only remove unreachable objects in run-time garbage-collected environments.

These approaches do not provide a complete run-time solution.

This dissertation provides two new approaches for memory leak detection and recovery in addition to a novel approach for dynamic memory management.

The first is memory leak detection (MLD) algorithm.

This algorithm reflects both of the physical and virtual behavior of memory allocation and benefits from the hardware support available for tracking physical pages in real memory in order to detect leak in virtual address space.

The latter is a memory leak detection and recovery (MLDR) algorithm based on a novel approach for dynamic memory management, a multi-layer virtual memory system (ML-VMS).

The ML-VMS reorganizes the currently used dynamic memory management and dynamic memory allocation mechanisms in order to solve or overcome the problem of memory leak The MLD uses a conservative approach to remove unreachable objects and save address space.

It delays possible application crashes due to the lack of virtual memory, but it can not solve the problem of stale objects.

The MLDR provides a thorough run-time solution.

It handles the problem of false positives, false negatives, and prevents target applications from crash due to the lack of virtual memory given a well-tuned parameters and that a target application can tolerate an additional overhead.

Both approaches are trace-driven simulated in order to provide a proof of concept and show the algorithms’ validity.

Our approach is compared to some current approaches for memory leak detection and recovery and is shown how it outperforms these approaches in providing a complete run-time solution.

The MLDR is recommended for mission critical applications that have to live for a long time and can tolerate a controllable overhead cost.

Key words: multi-layer virtual memory system, memory leak detection, memory leak recovery, virtual memory, memory aging, and dynamic memory allocation

Main Subjects

Mathematics

Topics

No. of Pages

169

Table of Contents

Table of contents.

Abstract.

Abstract in Arabic.

Chapter One : Introduction.

Chapter Two : Dynamic memory allocation and current approaches for solving memory leak problem.

Chapter Three : A new approach for memory leak detection (mld) using aging in physical memory space.

Chapter Four : Multi-layer virtual memory system (ml-vms).

Chapter Five : Performance evaluation and simulation.

Chapter Six : Conclusions and future work.

References.

American Psychological Association (APA)

Atum, Ahmad Ali. (2007). A novel multi-layer virtual memory system to solve memory leak problem. (Doctoral dissertations Theses and Dissertations Master). Amman Arab University, Jordan
https://search.emarefa.net/detail/BIM-528384

Modern Language Association (MLA)

Atum, Ahmad Ali. A novel multi-layer virtual memory system to solve memory leak problem. (Doctoral dissertations Theses and Dissertations Master). Amman Arab University. (2007).
https://search.emarefa.net/detail/BIM-528384

American Medical Association (AMA)

Atum, Ahmad Ali. (2007). A novel multi-layer virtual memory system to solve memory leak problem. (Doctoral dissertations Theses and Dissertations Master). Amman Arab University, Jordan
https://search.emarefa.net/detail/BIM-528384

Language

English

Data Type

Arab Theses

Record ID

BIM-528384