The overall power consumption of datacentres is increasing tremendously due to the high demand of... more The overall power consumption of datacentres is increasing tremendously due to the high demand of digital services. Moreover, the cooling load contributes up to 50% of the power consumption due to the higher densities of newer versions of servers. However, there is an increased awareness in the operations of the sub-systems, i.e. workload, cooling load and power consumption. This awareness of the interactions between the sub-systems provides a better understanding for maintaining the datacentre as an energy-efficient infrastructure. A direct contact liquid cooling technology is examined extensively by retrofitting to an air-cooled server. First the conventional SunFire V20z air-cooled server is benchmarked via SPECpower_ssj2008 workload to obtain some standard values. The server is placed inside a wind tunnel to ensure a controllable environment. Then an overall evaluation of the retrofitted server is presented and compared with the standard server. The retrofitted server shows a re...
The present work represents a two-dimensional numerical prediction of forced turbulent flow heat ... more The present work represents a two-dimensional numerical prediction of forced turbulent flow heat transfer through a grooved tube. Four geometric groove shapes (circular, rectangular, trapezoidal and triangular) were selected to perform the study, as well as two aspect ratios of groove-depth to tube diameter (e/D = 0.1 and 0.2). The study focuses on the influence of the geometrical shapes of grooves and groove-depth on heat transfer and fluid flow characteristics for Reynolds number ranging from 10,000 to 20,000. The characteristics of Nusselt number, friction factor and entropy generation are studied numerically by the aid of the computational fluid dynamics (CFD) commercial code of FLUENT. It is observed that the best performance occurs with the lower depth-groove ratio, whereas it is found that the grooved tube provides a considerable increase in heat transfer at about 64.4 % over the smooth tube and a maximum gain of 1.52 on thermal performance factor is obtained for the triangular groove with (e/D = 0.1).
Cloud datacenters are compute facilities formed by hundreds and thousands of heterogeneous server... more Cloud datacenters are compute facilities formed by hundreds and thousands of heterogeneous servers requiring significant power requirements to operate effectively. Servers are composed by multiple interacting subsystems including applications, microelectronic processors, and cooling which reflect their respective power profiles via different parameters. What is presently unknown is how to accurately model the holistic power usage of the entire server when including all these subsystems together. This becomes increasingly challenging when considering diverse utilization patterns, server hardware characteristics, air and liquid cooling techniques, and importantly quantifying the non-electrical energy cost imposed by cooling operation. Such a challenge arises due to the need for multidisciplinary expertise required to study server operation holistically. This work provides a unified model for capturing holistic power usage within Cloud datacenter servers. Constructed through controlled laboratory experiments, the model captures the relationship of server power usage between software, hardware, and cooling agnostic of architecture and cooling type (air and liquid). An exciting prospect is the ability to quantify the amount of non-electrical power consumed through cooling, allowing for more realistic and accurate server power profiles. This work represents the first empirically supported analysis and modeling of holistic power usage for Cloud datacenter servers, and bridges a significant gap between computer science and mechanical engineering research. Model validation through experiments demonstrates an average standard error of 3% for server power usage within both air and liquid cooled environments.
The natural convection cooling system of a fully immersed server in a dielectric liquid is analys... more The natural convection cooling system of a fully immersed server in a dielectric liquid is analysed numerically, where the servers are in sealed capsules and submerged in a dielectric fluid. A reduced order flow model is developed under a saturated porous media flow assumption using a Darcy flow regime and the Stokes equations solved numerically using successive over relaxation and time marching techniques. The simplified model is shown to agree well with predictions from full Navier–Stokes flow analyses and then used to study the role of spatial parameters on the convective heat transfer, in particular the effect of the locations and separations of two heat sources representing two central processing units (CPUs). The flow and heat transfer characteristics are analysed for a range of modified Rayleigh numbers between 0.5 and 300 and a correlation for Nusselt number is obtained which shows that the thermal behaviour is most strongly influenced by the modified Rayleigh number and that the vertical separation of the CPUs is more influential than the vertical position of the lower CPU.
The overall power consumption of datacentres is increasing tremendously due to the high demand of... more The overall power consumption of datacentres is increasing tremendously due to the high demand of digital services. Moreover, the cooling load contributes up to 50% of the power consumption due to the higher densities of newer versions of servers. However, there is an increased awareness in the operations of the sub-systems, i.e. workload, cooling load and power consumption. This awareness of the interactions between the sub-systems provides a better understanding for maintaining the datacentre as an energy-efficient infrastructure. A direct contact liquid cooling technology is examined extensively by retrofitting to an air-cooled server. First the conventional SunFire V20z air-cooled server is benchmarked via SPECpower_ssj2008 workload to obtain some standard values. The server is placed inside a wind tunnel to ensure a controllable environment. Then an overall evaluation of the retrofitted server is presented and compared with the standard server. The retrofitted server shows a re...
The present work represents a two-dimensional numerical prediction of forced turbulent flow heat ... more The present work represents a two-dimensional numerical prediction of forced turbulent flow heat transfer through a grooved tube. Four geometric groove shapes (circular, rectangular, trapezoidal and triangular) were selected to perform the study, as well as two aspect ratios of groove-depth to tube diameter (e/D = 0.1 and 0.2). The study focuses on the influence of the geometrical shapes of grooves and groove-depth on heat transfer and fluid flow characteristics for Reynolds number ranging from 10,000 to 20,000. The characteristics of Nusselt number, friction factor and entropy generation are studied numerically by the aid of the computational fluid dynamics (CFD) commercial code of FLUENT. It is observed that the best performance occurs with the lower depth-groove ratio, whereas it is found that the grooved tube provides a considerable increase in heat transfer at about 64.4 % over the smooth tube and a maximum gain of 1.52 on thermal performance factor is obtained for the triangular groove with (e/D = 0.1).
Cloud datacenters are compute facilities formed by hundreds and thousands of heterogeneous server... more Cloud datacenters are compute facilities formed by hundreds and thousands of heterogeneous servers requiring significant power requirements to operate effectively. Servers are composed by multiple interacting subsystems including applications, microelectronic processors, and cooling which reflect their respective power profiles via different parameters. What is presently unknown is how to accurately model the holistic power usage of the entire server when including all these subsystems together. This becomes increasingly challenging when considering diverse utilization patterns, server hardware characteristics, air and liquid cooling techniques, and importantly quantifying the non-electrical energy cost imposed by cooling operation. Such a challenge arises due to the need for multidisciplinary expertise required to study server operation holistically. This work provides a unified model for capturing holistic power usage within Cloud datacenter servers. Constructed through controlled laboratory experiments, the model captures the relationship of server power usage between software, hardware, and cooling agnostic of architecture and cooling type (air and liquid). An exciting prospect is the ability to quantify the amount of non-electrical power consumed through cooling, allowing for more realistic and accurate server power profiles. This work represents the first empirically supported analysis and modeling of holistic power usage for Cloud datacenter servers, and bridges a significant gap between computer science and mechanical engineering research. Model validation through experiments demonstrates an average standard error of 3% for server power usage within both air and liquid cooled environments.
The natural convection cooling system of a fully immersed server in a dielectric liquid is analys... more The natural convection cooling system of a fully immersed server in a dielectric liquid is analysed numerically, where the servers are in sealed capsules and submerged in a dielectric fluid. A reduced order flow model is developed under a saturated porous media flow assumption using a Darcy flow regime and the Stokes equations solved numerically using successive over relaxation and time marching techniques. The simplified model is shown to agree well with predictions from full Navier–Stokes flow analyses and then used to study the role of spatial parameters on the convective heat transfer, in particular the effect of the locations and separations of two heat sources representing two central processing units (CPUs). The flow and heat transfer characteristics are analysed for a range of modified Rayleigh numbers between 0.5 and 300 and a correlation for Nusselt number is obtained which shows that the thermal behaviour is most strongly influenced by the modified Rayleigh number and that the vertical separation of the CPUs is more influential than the vertical position of the lower CPU.
Uploads
Papers by Yaser Anii