Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Master of Science

Program

Computer Science

Supervisor

Hanan, Lutfiyya

Abstract

The concept of migration and checkpoint/restore has been a very important topic in research for many types of applications including any distributed systems/applications or single massive systems/applications; and low latency vehicular use cases, augmented reality(AR) and virtual reality(VR) applications. Migrating a service requires that the state of the service is preserved. This requires checkpointing the state and restoring it on a different server in multiple rounds to avoid a total loss of all data in case of a failure, fault or error. There are many different types of migration techniques utilized such as cold migration, pre-copy migration, post-copy migration.

Compared with the above migrations, MiGrror migration needs to consider rounds of memory changes for when it does the migration. The utilization of rounds of memory changes means that it is more precise to migrate instead of migrating on a inappropriate time interval. In this thesis, we will implement a testbed for the MiGrror and Pre-copy technique (as there are currently none that are not simulations) and evaluate their performance.

Summary for Lay Audience

In the recent years, there have been a massive push to use containers on the cloud rather than virtual machines. This increase has warranted new and more powerful resources in order to transfer a massive throughput through a very small amount of time that may be used on the cloud. Migration is the process of migrating information from one device to another.

A container is considered to be a more lightweight implementation of computing resource virtualization(containerization) compared to virtual machines. Container technology uses fewer computing resources than virtual machines, and can be deployed, expanded and migrated faster.

The process of migration is part of checkpoint and restore where the goal is to checkpoint application states and then restore them after so it is ready to be used later. Migration is mainly done in different types of techniques that differ by when/how the memory is transferred and the amount transferred. The main technique most containers use are a time gated one but might not be efficient enough for more low latency dependent applications.

Therefore, we implemented a testbed for a new proposed migration technique compared with the usual migration technique used by many containers. The implementation we proposed utilizes podman containers on two servers. We evaluated the metrics for both migration techniques migrating from one container to another. The results show that the proposed migration technique has a lower downtime for the migration compared with the main migration technique used.

Download

Share

COinS