P2P Cloud Storage System

Description

With the ever-increasing number of devices (mobile phones and sensors, servers, smart home appliances, etc) connecting to the Internet, we are confronted with many new challenges including data collection from distributed sources, data management in storage, query, synchronization, sharing and transfer, data analysis and data publishing. These lead to a new era in a computer science: big data. Currently, cloud computing gains competitive advantage to process the big data due to two distinct features: on-demand resource access that lowers the cost of infrastructure cost; high resource elasticity to meet the varying processing demands of data.

However, the nature of the data is widely distributed in the network. To enable the effective data analytics in the centralized cloud, there usually needs a large amount of data transfer to co-locate the data and computation which may reduce the performance and cost. In addition, some data might have legislation constraints on the locations and are amenable to be processed in the local. For these cases, we envision a decentralized P2P cloud storage among small data centers to overcome the limit of resource elasticity of individual data centers. Through collaboration with each other, each participant SDC can benefit in terms of performance and cost.

The p2p cloud storage poses new challenges and the prominent one is the heterogeneity in different small data centers such as performance heterogeneity in storage, network, and resource heterogeneity in availability, storage media. Due to those heterogeneity, it is difficult to guarantee the performance of data storage and transfer. Thus, it is important to define a new software-defined layer to translate the high-level performance goals to low-level resource requirements. This process may involve some machine learning algorithm to analyze the goals.

In this semester project, we will firstly implement a p2p cloud storage for different cloud vendors and design a layer which maps the service level agreements (SLAs) to the resource requirements. Specifically,

1) we will develop and deploy a p2p cloud stroage service in a heterogeneous environment and conduct performance evaluation in different locations.

2) To deal with the heterogeneity, we will design a new layer to map the high level SLAs to low-level resource requirement. This can involve the design, implementation, test and documentation.

• Having the motivation for indulging in a research oriented project
• Programming skills with Java
• Interested in algorithm design

Contacts

In case of any questions, please drop us an email or come to our offices:

• Hao Zhuang (BC 118)
• Rameez Rahman (BC 142)