Microsoft Technology Licensing, LLC

Scott Thomas Seaton

Allan Joseph Wenzel

Keith Allen Krueger

Christian L Belady

David Thomas Gauthier

Osvaldo Patricio Morales

Stephan W Gilges

Lalu Vannankandy Kunnath

Daniel Gerard Costello

Kris Sean Redding

Brian Andersen

Peter Odongo Wangia

Cheerei Cheng

Jens Conrad Housley

Steven Ray Steinert

Steve Todd Solomon

Ryan Douglas Carver

Brian Jon Mattson

Kenneth Allen Lundgren

Jose Seah Delapaz

Eric Matthew Beal

Shaun L Harris

Glen Robert Beyer

Kristofer Andrew Johnson

Mark Joseph Baracani

Steven Solomon

Courtney Shepherd Jones

Jayanthi Lakshmanan

Yidian Sun

Daniel David Woodman

Robert G Allison

Daniel N McManus

Joseph L Soroka

Vincent Karczewski

Shu-Kuen Raymond Chiu

A design for the electrical infrastructure of a data center enables two different configurations to be utilized, including a distributed, redundant configuration that provides higher reliability and a non-redundant configuration that provides higher capacity. In the distributed, redundant configuration, each server in the data center draws power from at least two different power supply systems. This enables load shifting when a power supply system becomes unavailable, which can have the effect of increasing server reliability. In the non-redundant configuration, each server in the data center draws power from only one power supply system. Load shifting is not utilized in the non-redundant configuration, which eliminates the need to maintain reserve capacity and thereby increases capacity. Advantageously, it is possible to switch between these two configurations without making any internal changes to the data center other than modifying connections between sets of server racks and power buses.

Variable reliability and capacity data center design

A first power train that includes a first plurality of components, and a second power train includes a second plurality of components. The first power train is configured to provide power to a first plurality of server racks of a first data center at a first level of high-availability service associated with a first uptime. The first plurality of components includes a first subset of the first plurality of components and a second subset of the first plurality of components. The second power train is configured to provide power to a second plurality of server racks of the first data center at a second level of high-availability service that is associated with a second uptime that is less than the first uptime. The second plurality of components includes a first subset of the second plurality of components and the second subset of the first plurality of components.

Variable redundancy data center power topology

Methods, systems, and apparatuses are described that enable the recovery of stranded power in a data center. For example, a power distribution system for recovering stranded power in a data center includes a first power distribution unit (PDU), a first busway segment that is operable to electrically connect the first PDU to a first set of server racks in a first row of server racks, a second busway segment that is operable to electrically connect the first PDU to a second set of server racks in a second row of server racks, a second PDU, a third busway segment that is operable to electrically connect the second PDU to a third set of server racks in a third row of server racks, and a fourth busway segment that is operable to electrically connect the second PDU to a fourth set of server racks in the second row of server racks.

Data center stranded power recovery design

In one example, a data center may be built in modular components that may be pre-manufactured and separately deployable. Each modular component may provide functionality such as server capacity, cooling capacity, fire protection, resistance to electrical failure. Some components may be added to the data center by connecting them to the center's utility spine, and others may be added by connecting them to other components. The spine itself may be a modular component, so that spine capacity can be expanded or contracted by adding or removing spine modules. The various components may implement functions that are part of standards for various levels of reliability for data centers. Thus, the reliability level that a data center meets may be increased or decreased to fit the circumstances by adding or removing components.

Modularization of data center functions

Key interlock systems and methods are described for safely carrying out a closed-transition procedure in an electric power distribution system in which two load busses that can be separately powered by the same power source or by two different power sources can be connected together via one or more bus tie breakers and in which a static transfer switch is used to selectively deliver power from one of the two different power sources to at least one of the load busses. Embodiments described herein prohibit access to a key that is required to close a bus tie breaker that connects the two load busses until at least a determination is made that a particular bypass breaker of the static transfer switch has been closed, thereby ensuring that both load busses are connected to the same power source.

Key interlock system and method for safe operation of electric power distribution system

A power distribution system among a set of units (e.g., server blocks) may comprise, for each unit, a utility line and a unit generator, and a reserve generator providing failover transient performance and redundancy improvement to the power for the unit generators. The reserve generator may connect to the units via a reserve bus, and the unit generators may selectively connect to a wraparound bus connected to the reserve bus. When the failover load exceeds the available failover transient capability of one generator, one or more unit generators may (automatically or by operator selection) be connected with the wraparound bus to apply available transient capability to satisfy the excess failover load with minimal increase in power distribution resources and complexity.

Power distribution with wraparound bus

Super capacitor supplemented server power is described. In embodiments, a power system manager is implemented to monitor the capability of one or more power supplies to provide power for a server system. The power system manager can determine that the capability of the power supplies to provide the power is deficient, and then engage one or more super capacitor power modules to provide supplemental power for the server system to mitigate the power deficiency.

Super capacitor supplemented server power

Machines may be connected to power distribution units (PDU/STSs) in such a way that if one PDU/STS fails or is taken out of service, the machines' loads are shifted more-or-less evenly to the other PDU/STSs. Several PDU/STSs may be provided for a group of dual-corded machines. The two cords of each machine are connected to a pair of PDU/STSs. Different pairs of PDU/STSs may be used for different machines. In normal operation, each machine may draw part of its power through each of the two cords. If one PDU/STS to which a machine is connected fails or is taken out of service, the entire load on that machine may be shifted to its remaining cord. Since redistribution may be approximately even across PDU/STSs, tolerance for one PDU/STS failure can be implemented with relatively little over-sizing of the PDU/STSs.

Connection between machines and power source

Growing community of inventors

Kirkland, WA, United States of America

Scott Thomas Seaton