Project offerings 2011

Projects supervised by Albert Zomaya

Virtualisation and Beyond for Greening IT
The major technological shift being witnessed particularly in the past few years is the advocacy of sustainability (more specifically, green IT in our case) from high performance. The field of green IT has already become a major research and development priority not just in academia and industry (e.g., Intel, IBM, Microsoft, Google, Amazon and VMWare). Many countries including Australia also have initiated numerous national/international investigation and development efforts to make IT more energy efficient. At first glance, this is related to the development of low-power and energy-efficient hardware technologies. However, greening IT goes beyond advancements of hardware. Software approaches like energy-aware scheduling and resource allocation can significantly reduce environmental footprints (e.g., energy consumption and carbon emissions). There are multiple levels in computer systems we can apply software approaches to including platforms, operating systems, virtual machines and service-level management. In this project, we will investigate various software-based energy-saving techniques primarily focusing on virtual machine monitoring and service-level workload management. Students involved in this project will have a chance to conduct their investigation and development using virtualization tools (e.g., VMWare products, Xen hypervisor and Sun Microsystems VirtualBox) in real multiprocessor computing systems. The resulting methodology of this project can have an immediate real-world impact.

Students with good programming background particularly in Unix-like environments are encouraged to apply.

Investigate Scheduling Policies inside a Virtual Machine Monitor
Virtualization is an important technology for improving datacentre efficiency. With virtualization, applications run on an abstract machine, or virtual machine (VM) implemented by software in a similar way of running on a real machine. The efficiency is achieved through consolidating a few virtual machines to a physical machine so that the utilization of the physical machine can be improved. The project intends to investigate the CPU scheduling policies used in a virtual machine monitor (VMM) and the impact to the performance of applications. The project also intends to implement a customized processor-sharing scheduling policy inside Xen Hypervisor to give each VM fair share of CPU cycles.

The student is expected to be comfortable with code reading and C/C++ programming in a Linux environment.

Comparison of Market-based and Best-effort Based Resource Allocation in Clouds
Virtualization and decoupling resources used by applications are important means that enable a cloud computing platform to efficiently accommodate a variety of client applications. CPU cycles, storage and network bandwidth can therefore be provisioned separately to heterogeneous applications. There are mainly two methods to allocate these resources to applications: one is best-effort based and the other is market-based. The best-effort method dynamically allocates resources whenever an application requests them, while the market-based method prices resources for different time slices according to the pre-specified demand from applications. The project intends to investigate via simulation the efficacy of the two methods from both applications’ and resource providers’ perspective. Particularly the project intends to give insight into the role of coordination of the resource allocation schemes in the context that decoupled resources are allocated separately in clouds.

Link Service Level Agreement (SLA) to Cloud Computing Energy Efficiency
The cloud computing paradigm enables multiple tenants to share a set of physical resources. It is of a cloud infrastructure provider’s interest to make efficient use of these physical resources in terms of cutting off its energy bill. The tenants, on the other hand, are interested in the quality of service they receive, which means that a tenant expects a resource is available when needed. The efficiency requirement from the provider often conflicts with the resource availability requirements from the tenants. The confliction can be reconciled through reaching a Service Level Agreement (SLA). However, there is a gap between SLA metrics and the underlying mechanisms for achieving efficiency. The project will examine common SLA metrics and investigate how to link SLA metrics to efficient resource allocation schemes. The outcome of the project will be a realistic model evaluated in a cloud computing platform and/or a few SLA-aware resource allocation algorithms based on the model.

The student is expected to have an understanding of operating systems as well as have basic performance modelling knowledge.

An Efficient Scheduling Scheme for Vehicular Ad hoc Networks
The purpose of Intelligent Transport System (ITS) is to provide efficient network services to vehicle drivers. These services can be entertainment (video and voice) information, map based guidance, emergency information etc. Vehicular Ad hoc Networks (VANETS) provide networking environment for ITS. VANETs are special type of mobile ad hoc wireless networks where vehicles act as nodes. As VANETs are becoming popular more and more people want to access and share data from their vehicles. There are two modes for accessing data, one is through vehicle to vehicle and the other is through roadside units (RSUs). This project investigates later one. When there are many vehicles in the system, and they want to access data through a RSU, scheduling of data/services becomes an important issue. The purpose of this project is to present an efficient scheduling scheme for road side units (RSUs) in VANETs. The algorithm will assign weights to requests based on the speed and distance of the vehicle from the RSU. A time stamp called earliest deadline first (EDF) is calculated for each request. It will depend on the speed and distance of the vehicles in a certain area around RSU called zone. The vehicles will be served in the form of groups or classes in these zones. Vehicles requiring same type of service can be served with the same weight. High speed and short distance from the RSU is assigned more weight and vice versa. This weight will change dynamically depending on the situation at a particular moment. In normal conditions services are treated equally, but as soon as an emergency situation arises more weight is assigned to that service, other services will be treated as best effort.

The student is expected to have good programming skills and networks knowledge.