Học bổng Tiến sĩ cho sinh viên UIT của Công ty Redhat

Cty RedHat (Hoa Kỳ) sẽ tài trợ cho nghiên cứu sinh của UIT. File đăng ký đính kèm theo topic này.

Thời hạn nộp phiếu proposal (bản cứng) về Phòng Quan hệ Đối ngoại: 19/10/2015 (Thứ Hai). Người nhận phiếu: Ms. Thùy Linh. 

Một số thông tin liên quan


PhD Research topic and Abstract:

Tamper resistant event logging in the cloud The legal implications of data and applications being held by a third party are complex and not well understood. If something goes wrong (say data is lost, or computation returns erroneous results), how do we determine who caused the problem (the customer or the provider)? In the absence of solid evidence, it would be impossible to settle disputes. The project will investigate the feasibility of constructing a tamper‐resistant, trustworthy log that records service interaction events at sufficiently fine granularity to enable investigations of accidents or incidents.

Logging should have the following characteristics:

*logging must guarantee fairness and non‐repudiation, ensuring that well‐behaved parties are not disadvantaged by the misbehaviour of others and that no party can subsequently deny their participation.

* it should enable tracing back the causes of an 'incident' (behaviour that is not contract compliant, that is, deviation from expected/legitimate behaviour) after it has occurred;

* its presence must not limit the functioning and the types of services offered by the cloud itself;

* logging itself should be constructed using cloud computing services.

It is an open question at the moment whether such logging is feasible. For this reason, The project will create the scientific foundations of secure logging in a potentially hostile environment.


Efficient and practical support for replication management in managed hosting environments The proposed research will develop practical and cost-effective solutions to a fundamental problem in crash-tolerant replication in managed hosting environments, e.g., cloud computing and data centres. These solutions will enable replicas to observe events in an identical order. The major challenge to protocol simplicity and cost-effectiveness arises from having to resolve a basic question: how long is ‘too long’ before a crash can be ascertained. The known approaches either strictly limit the maximum load the system can ever be subject to and thus address this question by defining a concrete upper bound, or develop complex solutions that never compromise correctness even when an assumed bound fails to hold but extract a huge overhead. Our approach will predict, in probabilistic terms, the delays to unfold in near future based on the delay measurements taken in the recent past. It will thus address the basic question in a probabilistic manner and develop solutions that offer a meaningful trade-off between correctness probability and overhead. We envisage that, at a moderate cost, the probability of an incorrect ordering can be made negligibly small in managed hosting environments .