Christian L. Belady

Microsoft · Datacenter Strategy, Planning and Development

A datacenter may use heat collected from a heat exchanger at the exhaust portion of a cooling system to heat inlet louvers for an atmospheric intake. The louvers may have fluid passages through which heated fluid may pass and cause the louvers to heat up. The heated louvers may operate during periods of snow, rain, high humidity, or other condition...

An exemplary method may include thermally coupling a network of cooling lines to each of a plurality of heat exchangers in a cooling system. The method may also include providing a first connection from a network of cooling lines to a first fluid source and a second connection from the network of cooling lines to a second fluid source for configuri...

A chassis includes a component interconnect board having multiple multi-function slots coupled thereto. Each of the multi-function slots is configured to accept different types of modules, including a battery module and one or more other types of modules, such as server modules or storage modules. The component interconnect board also includes a po...

A data center system is described which includes multiple data centers powered by multiple power sources, including any combination of renewable power sources and on-grid utility power sources. The data center system also includes a management system for managing execution of computational tasks by moving data components associated with the computa...

Pre-computing a portion of forecasted workloads may enable load-balancing of data center workload, which may ultimately reduce capital and operational costs associated with data centers. Computing tasks performed by the data centers may be analyzed to identify computing tasks that are eligible for pre-computing, and may be performed prior to an act...

Resource-based adaptive server loading is described. In embodiments, a current load level can be determined for a resource that is utilized by an adaptive server system to process computer-executable instructions that are a dynamic computational demand on the adaptive server system. The current load level is compared with a target load level for th...

This document describes various techniques for automatically starting servers at low temperatures. A server may be powered on into a heating mode responsive to determining that a temperature of the server is below an operational temperature range. The server may then be restarted when the temperature of the server has increased to a temperature tha...

A flexible midplane comprises: a printed circuit board including: a middle section and at least one side section; and a flexible region disposed between the middle section and each side section; wherein each flexible region permits the corresponding side section to be bent in relation to the middle section.

A server computer may be located at a home or business and may be used for heating the home or business. The server computer may operate a distributed cloud based application and may generate heat in response to application load. The server computer may have an air or water based heat exchanger that may integrate into various heating, ventilation,...

Power is provided to one or more devices in a system that includes a hierarchical power smoothing environment having multiple tiers. In response to a peak in power usage by the one or more devices, power is provided from a first power smoothing component in a first tier of the multiple tiers. Additionally, power is provided to the one or more devic...

An electronic device has a ventilation system with an inlet section that receives inlet air that travels past components of the computing device to be cooled and exits at an outlet section. The air carries heat away from the components. A liquid heat transfer system captures heat generated by the components and transfers the captured heat to the in...

Computer processes are allocated among cores of a multi-core multi-processor system at least in part as a function of utilization histories of the processes.

Redundant cooling systems and methods are disclosed. In an exemplary embodiment, a method for redundant cooling system of computer systems and other electronics may comprise thermally connecting a cooling fluid to one or more heat-generating components to absorb heat from the heat-generating components during operation. The method may also comprise...

A rack power unit is configured to be inserted into a device rack of a data center. The rack power unit includes one or more power supplies and one or more battery packs. The one or more power supplies are each configured to receive power (e.g., AC power) when the apparatus is in the device rack, and convert the received power to a DC power. The on...

The promise of cloud computing is, nowadays, mostly limited to the developed regions of the world where, approximately, only one half of the world's population lives. In this paper, we present an attempt to bring the cloud to the majority populations of the developing world, with the help of long-distance, wirelessly connected and renewable-energy-...

In this paper, we argue that servers can be sent to homes and office buildings and used as a primary heat source. We call this approach the Data Furnace or DF. Data Furances have three advantages over traditional data centers: 1) a smaller carbon footprint 2) reduced total cost of ownership per server 3) closer proximity to the users. From the home...

Data centers are the heart of the global economy. In the mid-1990s, the costs of these large computing facilities were dominated by the costs of the information technology (IT) equipment that they housed, but no longer. As the electrical power used by IT equipment per dollar of equipment cost has increased, the annualized facility costs associated...

This paper provides several metrics to characterize the efficiency of data centers. Power Usage Effectiveness (PUE) measures the fraction of the total facility power devoted to IT work. Compute Power Efficiency (CPE) measures the overall efficiency of the data center, considering both power and cooling and the utilization of the data center IT equi...

This work describes two data center efficiency metrics: Power Usage Effectiveness (PUE) and Compute Power Efficiency (CPE). PUE characterizes the fraction of the total data center power used for IT work. CPE characterizes the overall data center efficiency, considering IT equipment utilization as well as how power is used in the data center. The PU...

This paper introduces three metrics, PUE, DCE and EAC, that aid in providing an objective view of the true state of data center operations with respect to power usage. The Power Usage Effectiveness (PUE) metric characterizes the utilization of the total power available for a data center. Lowering PUE by improving the efficiency of the cooling solut...

As power and heat densities continue to increase, the behavior of the data-center environment grows in importance, especially when developing new in-room cooling technologies. As a result, there has been a continued interest in understanding the behavior of various technologies in data centers. One issue is whether gravity plays a role in air coole...

With the densification of electronics, predictions for electronics power trends have been closely watched. For semiconductors, the ITRS (International Technology Roadmap for Semiconductors) has published power trends for processors for years. For electronic equipment such as servers, Uptime published a power trend in 2000 and more recently in 2004,...

Computer manufacturer’s are constantly trying to tweek more performance out of their existing products by using the highest performing processors. Typically, manufacturers upgrade the platforms by simply replacing the old processor with the latest speed processor. Like other manufacturers, HP generally follows this practice with the exception ot HP...

This presentation gives a brief perspective on the cooling challenges that manufacturers and their customers facing at the chip level, box level, rack level and data center level. In particular, the goal of the discussion is to demonstrate that high density packaging is not a problem that is unique to chip level cooling but rather the whole value c...

Today’s data centers are designed for handling heat densities of 1000W/m2 at the room level. Trends indicate that these heat densities will exceed 3000W/m2 in the near future. As a result, cooling of data centers has emerged as an area of increasing importance in electronics thermal management. With these high heat loads, data center layout and des...

This paper presents the thermal design methodology used to design a multi-processor enterprise server, the RP8400. The proposed methodology combines well-known analytical and experimental thermal design tools, including heat transfer correlations, Flow Network Modeling (FNM) and Computational Fluid Dynamics (CFD) techniques, and experimental measur...

Recently technology advances have driven the power densities of electronics to unprecedented levels, but unfortunately central office and data center infrastructure technology advancements have not kept pace. Trends indicate that room densities for both central offices and data centers will jump from less than 1000 W/m² to densities of...

With semiconductor power increasing rapidly, there is no doubt that package level cooling and infrastructure level power and cooling issues are important engineering problems to solve. Some of the relevant issues are discussed in this paper

A low noise fan reduces the production of turbulent vortices created by the passage of the fan's blade tips through the air by providing an annular venturi that is attached at its inner surface to the blade tips, and that thus rotates with those blades as a unit. This prevents high pressure air at the blade tips from spilling into low pressure air....

This paper outlines the technique of Flow Network Modeling (FNM) for system-level thermal design of electronic devices. FNM is a generalized methodology involving representation of a cooling system as a network of components and flow paths for predicting systemwide distribution of flow rates, pressures, and temperatures. Performance of the individu...

The evolution of the microprocessor has led to a significant increase in the number of I/O pads, the number of high frequency interconnects and power dissipation. The increase in performance has reduced board to board spacing, and as a result, the available space for thermal design in a computer system. High power dissipation and the small form fac...

The ability to accurately characterize heat sinks is becoming more important in the electronics industry as design margins decrease and systems become more complex. Although heat sinks can be characterized many ways, there are limitations associated with each. The three most common characterization techniques are empirical and theoretical correlati...

This paper provides an overview of the effects of altitude on electronics cooling. MIL-STD 210 is reviewed to demonstrate the relationship of altitude to density as well as to outside air temperature. Once these relationships are understood the paper discusses the impact of altitude on the performance of air moving devices using the widely accepted...

High power dissipation from microprocessors, support chips, memory chips and mass storage has resulted in large overall power dissipation from computer systems. The deployment of these computer systems in large numbers and in very dense configurations in a data center has resulted in very high power densities at room level. These computer systems a...