The Computing Infrastructure Research Centre (CIRC) at McMaster University is bringing transformative changes in the way data centres (DCs) and other forms of computing infrastructure are designed, built, and operated. CIRC is the first data centre research facility in Canada and among very few others worldwide, and boasts a pioneering model of university-industry collaboration. Its research is market-focused and needs-driven, ensuring that its resources are fully leveraged in solving problems that create value to industry and society. CIRC is looking for a PhD Students in the following areas,
1 .Data Centre Load Profile Modelling
Identifying load trends generally depends on the analysis of operational data and the type, availability and capacity of the Data Centre (DC) being employed. Incoming workloads can then be classified and allocated to corresponding servers and predictive algorithms can be developed to profile the various DC types in association with the workloads. The goal of this project is to develop a tool to enable the development of DC load profiles for different DC types. This information can in turn be used to forecast usage and optimize online capacity. The development of a family of load profile scenarios, will provide visibility to potential variations in configuration of a DC and its impact on such metrics as Total Cost of Operation.
2. Data Centre Design, Operations and Maintenance Decision Tool
In this research student will work on developing individual models for power, IT, and cooling equipment. A decision tool will be designed that incorporates the above models, optimizing equipment selection and capacity. The influence of data centre operational settings and configuration on the total cost of operation (TCO) will also be investigated. Finally, a maintenance decision tool will be designed to analyze the operational data to predict the influence of a maintenance activity on the TCO of a facility, enabling the development of appropriate maintenance schedules.
3. Optimization of Modular Data Centres
The research involves development of rapid design optimization methodologies for modular data centres (MDCs). Due to a large number of possible choices of each modular component, identifying an optimal design for MDC is nontrivial, especially, when the individual components have significant trade-off in overall system performance. The first focal point of the research will be to embed computationally efficient MDC models that describe the trade-offs in the system into a combinatorial optimization problem. The presence of complex discrete search/parameter spaces and functions in the optimization problem rules out the use of calculus based solution techniques. Thus, the second focal point will be exploration of efficient solution techniques.
4. Design and Development of Single-stage Power Unit for Data Centre Applications
In order to achieve high overall efficiency, all power conversion stages in the power distribution system need to be operated and realized with the highest possible efficiency. In general, two types of power converters can be employed in the power supply unit, namely AC-DC converters and DC-DC converters. These converters have fundamentally different properties, and in this project, the available AC-DC and DC-DC power converter topologies will be reviewed and their characteristics will be studied. Based on observations from that review, a new power unit will be designed and developed to address the efficiency issue while improving the power density, power quality and cost of the power unit.
5. Design and Development of Supervisory Control and Energy Management in Data Centres
To improve the efficiency of (modular) DCs at different operating conditions, and particularly at light load conditions, a parallel structure for the power supplies will be proposed. This structure provides the opportunity to operate each power conversion unit close to its maximum efficiency conditions. The parallel structure can appear at both the UPS and PSU units. Parallel structure at UPS unit means using number of AC-DC converters in parallel. In this case there should be a supervisory controller that investigates the power demand and selects the number of AC-DC converter that should be operating. Currently, UPS systems are often constructed using several small inverters instead of one large inverter. In that configuration, the number of inverters is limited by the chassis controlling them. Removing that, along with the development of a new control algorithms would release this constraint. The ultimate goal for the new configuration is to add communication mechanisms to inverters, and thus enable a plug and play configuration for inverters.
We are looking for a candidate who holds a Masters degree with an excellent academic record in Computer Science, Computer Engineering, Electrical Engineering, or related fields from internationally recognized Universities. A strong background in mathematical optimization is desirable. For further information about the position, please contact us at “firstname.lastname@example.org”.
1. Workload Profiling for Edge Data Centres
The goal of this project is to develop workload models for different types of servers, including HPC, cloud, and edge data centers. The candidate will also deploy an open source resource management tool on our in house date center so that the different workload models can be developed on this framework and different resource management algorithms can be tested.
We are looking for a candidate who has an excellent academic record in Computer Science, Electrical and Computer Engineering, or related fields from internationally recognized universities. Experience with statistical modeling and data analysis is a must. A strong interest in experimental work is required. For further information about the position, please contact Dr. Douglas Down (email@example.com) and/or visit our website circ.mcmaster.ca