Superconductor networks for dynamically reconfigurable power planes

Abstract

The present application relates to a superconductor network for a dynamically reconfigurable power plane. A system comprising: a superconducting cable network configured to supply power to a plurality of server racks arranged in a space of a building; a first power connection configured to connect a first power source from a first side of the building to the superconducting cable network and configured to supply power to a first section of the superconducting cable network; a second power connection configured to connect a second power source from a second side of the building different from the first side to the superconducting cable network and configured to supply power to a second section of the superconducting cable network; and a plurality of bus ducts, each bus duct configured to connect the superconducting cable network to one or more of the plurality of server racks.

Description

Technical Field

[0001] The present disclosure relates to superconductor networks for dynamically reconfigurable power planes. Background Art

[0002] Large-scale data centers include many server racks that require power to operate. The data center building may have several floor levels, each with server racks on it. For each floor level, the server racks may be kept on the floor while power is routed to the server racks via bus ducts located on the floor level above, below, or beside the server racks. The power provided to a single floor level of a data center may be on the order of tens of megawatts, and the power provided to the building of the data center may be on the order of hundreds of megawatts.

[0003] Power is provided from one or more power sources. Typically, power is routed into the building from one side and divided between aisles of server racks via a branching network of bus ducts. In an example system, power can be fed from one or more power sources to one or more floor levels of a data center via split electrical connections, and the electrical connections are connected to parallel bus ducts. The bus ducts deliver power from a first side of the building to a second side of the building. Each bus duct can be connected to a server rack along parallel aisles within the building.

[0004] Power distribution suffers from several disadvantages. Because power is only provided from the first side of the building, the current density of the power provided is much higher on the first side than on the second side. However, bus ducts are typically designed uniformly, which results in underutilization of the conductive material of the bus ducts closer to the second side of the building. Complex and expensive infrastructure, including transfer switches and circuit breakers, is also required to shunt power to each bus duct and to each server rack in multiple aisles of the data center.

[0005] Another problem with one-way power distribution is that failed upstream equipment can cause downstream equipment to shut down. Redundant bus ducts and additional infrastructure are needed to provide backup distribution solutions when upstream equipment fails, and to accommodate maintenance of one section of the system without taking the rest of the system offline.

[0006] Additionally, because each aisle of a data center is designed to receive the same amount of power, it is necessary to balance the data center's server racks between bus ducts and aisles so that the power density of one bus duct or aisle is not unbalanced with the power density of another bus duct or aisle. To avoid imbalance, servers that require relatively high current density, such as machine learning data catalog (MLDC) servers, must be distributed throughout the data center, and careful advance planning is required to avoid concentrating these high power density servers in a single part of the building. Summary of the invention

[0007] The present disclosure provides a superconducting cable network included in a dynamically configurable power plane for powering large-scale systems such as data centers. The power plane can be arranged to receive power from multiple sides of a building, and the superconducting cables can control the direction and distribution of the received power throughout the building.

[0008] One aspect of the present disclosure relates to a system comprising: a superconducting cable network configured to supply power to a plurality of server racks arranged within a space of a building; a first power connection configured to connect a first power source from a first side of the building to the superconducting cable network and configured to supply power to a first section of the superconducting cable network; a second power connection configured to connect a second power source from a second side of the building different from the first side to the superconducting cable network and configured to supply power to a second section of the superconducting cable network; and a plurality of bus ducts, each bus duct configured to connect the superconducting cable network to one or more of the plurality of server racks.

[0009] In some examples, the first side and the second side may be opposite sides of a building.

[0010] In some examples, the system may further include a plurality of switch controllers. Each switch controller may be configured to control an electrical connection between at least one of the first power connection and the second power connection and at least one server rack.

[0011] In some examples, the plurality of switch controllers may include a superconducting switch positioned between the first section and the second section of the superconducting cable network and may be configured to control an electrical connection between the first section and the second section of the superconducting cable network.

[0012] In some examples, the plurality of switch controllers may include an electrical switch positioned at the first bus duct and may be configured to control electrical connections between the first bus duct and one or more server racks connected to the first bus duct.

[0013] In some examples, the plurality of switch controllers may include an electrical switch positioned at the first server rack and may be configured to control electrical connections between a plurality of bus ducts connected to the first server rack and the first server rack.

[0014] In some examples, the plurality of server racks may be divided into a plurality of discrete sections, wherein each discrete section is connected to each of a first power source and a second power source. The plurality of switch controllers may be configured to disconnect any one of the discrete sections of the plurality of server racks from the first power source and the second power source while other sections of the plurality of server racks remain connected to at least one of the first power source or the second power source.

[0015] In some examples, a network of superconducting cables may be arranged to allow electricity to flow therethrough in a serpentine configuration.

[0016] In some examples, the network of superconducting cables may be arranged to allow power to flow therethrough in a toroidal configuration.

[0017] In some examples, a network of superconducting cables may be arranged to allow power to flow therethrough in a mesh configuration.

[0018] In some examples, the superconducting cable network may be positioned directly above the plurality of server racks.

[0019] In some examples, a space in a building may include a first section including a first portion of servers assigned a relatively higher priority and a second section including a second portion of servers assigned a relatively lower priority.

[0020] In some examples, the first portion of servers may include at least one of a machine learning server or a network device.

[0021] In some examples, the superconducting cable may be configured to transmit a direct current (DC) voltage between approximately 1 kV and 40 kV.

[0022] In some examples, the superconducting cable can be configured to send a DC current between approximately 10 kA and 30 kA.

[0023] Another aspect of the present disclosure relates to a data center, comprising: a system according to any one of the embodiments described herein; and a plurality of server racks arranged within a space of a building.

[0024] Another aspect of the present disclosure relates to a server maintenance method for updating a portion of servers in a data center, the data center comprising a superconducting cable network configured to supply power from each of a first power source and a second power source to a plurality of server racks arranged in a space of a building. The superconducting cable network can be connected to the first power source via a first power connection positioned on a first side of the building and connected to the second power source via a second power connection positioned on a second side of the building different from the first side. One or more controllers can be configured to control the electrical connection between the first power source and the second power source and the plurality of server racks. The method may include: providing a first control signal from the one or more controllers to electrically disconnect the first portion of the server from the first power supply and the second power supply without disconnecting the second portion of the server from the first power supply and the second power supply; performing maintenance on the first portion of the server; providing a second control signal from the one or more controllers to electrically reconnect the first portion of the server to the first power supply and the second power supply without disconnecting the second portion of the server from the first power supply and the second power supply; providing a third control signal from the one or more controllers to electrically disconnect the second portion of the server from the first power supply and the second power supply without disconnecting the first portion of the server from the first power supply and the second power supply; performing maintenance on the second portion of the server; and providing a fourth control signal from the one or more controllers to electrically reconnect the second portion of the server to the first power supply and the second power supply without disconnecting the first portion of the server from the first power supply and the second power supply. The first portion of the server and the second portion of the server may be divided or segmented according to corresponding confidentiality levels.

[0025] In some examples, a first control signal may electrically disconnect a first section of the superconducting cable network from a second section of the superconducting cable network, and a second control signal may electrically reconnect the first section of the superconducting cable network to the second section of the superconducting cable network.

[0026] In some examples, a first control signal may electrically disconnect the first portion of servers from the superconducting cable network at one or more DC bus conduits configured to provide power from the superconducting cable network to the first portion of servers, and a second control signal may electrically reconnect the first portion of servers to the superconducting cable network at the one or more DC bus conduits.

[0027] In some examples, a first control signal can electrically disconnect a first server rack of a first portion of servers from each of a first DC bus conduit and a second DC bus conduit, wherein the first DC bus conduit is configured to provide power to the first server rack from a first power supply and the second DC bus conduit is configured to provide power to the first server rack from a second power supply, and a second control signal can electrically reconnect the first server rack to the first DC bus conduit and the second DC bus conduit. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a schematic diagram of electrical connections for an example data center system according to aspects of the present disclosure.

[0029] Figure 2 is a schematic diagram of example electrical connections between a power plane and server infrastructure according to aspects of the present disclosure.

[0030] Figure 3 and Figure 4 According to various aspects of the present disclosure Figure 2 Example implementations of example electrical connections.

[0031] Figure 5 is a schematic diagram of another example electrical connection between a power plane and server infrastructure according to aspects of the present disclosure.

[0032] Figure 6 According to various aspects of the present disclosure Figure 6 Example implementations of example electrical connections.

[0033] Figure 7 is a schematic diagram of another example electrical connection between a power plane and server infrastructure according to aspects of the present disclosure.

[0034] Figure 8 and Fig. 9 According to various aspects of the present disclosure Figure 8 Example implementation of example electrical connections

[0035] Fig.10 is a flow diagram of an example server maintenance routine in accordance with aspects of the present disclosure. DETAILED DESCRIPTION

[0036] Overview

[0037] The technology generally relates to a superconductor cable network for distributing power in a large-scale system such as a data center, and more particularly to a switching network of a dynamically configurable power plane for powering a large-scale system such as a data center.

[0038] Power may be distributed to large-scale systems via a power bus. The power bus may include a bus duct for carrying power. The power bus may include several connection elements, including a bus duct connection for connecting the power bus to rows of server racks in a data center, and power connections for connecting the power bus to one or more power sources at multiple points. The switch network may include a plurality of switches included in the power bus and configured to direct power between the power sources and the bus duct connection to provide power to all of the server racks.

[0039] Control signals may be used to control switches in order to distribute power throughout the data center in a desired manner. For example, power may be diverted from a section of the data center while the remainder of the data center continues to receive power and remain operational. This may be done to perform maintenance on a disconnected section of the data center while the rest of the data center remains operational.

[0040] The system can be arranged to control the concentration of power within the power bus. For example, power connections can be located in different quadrants or sides of the data center. Additionally, bidirectional and multi-directional switches can be provided within the power bus to facilitate redirecting power throughout the system. Such a configuration avoids the unidirectional flow of current from the high current density side of the data center to the low current density side. Therefore, it may be possible to achieve a more uniform current density throughout the data center while enabling high power density to be directed to specific locations in the data center as needed.

[0041] The arrangements described herein provide a number of advantages over conventional data center power buses that provide unidirectional current flow from "upstream" busbar elements to "downstream" elements. For example, conventional data center power buses may require balancing between server racks at different current densities, have the risk of upstream faults causing downstream power outages, and generally underutilize downstream busbar elements because current density decreases with increasing distance from the power source. In contrast, the power bus and switch network arrangements of the present disclosure allow for bidirectional and multidirectional current flow, meaning that busbar elements are not "upstream" or "downstream" of each other, current density can be balanced throughout the data center, and busbar elements are not underutilized.

[0042] Example System

[0043] Figure 1 An example system 100 is illustrated for distributing power in a large-scale data center 110. The data center 110 may receive power from more than one side of a building. For example, a first side 112 of the building may include an electrical connection between at least a first power source 122 and a power bus 130, while a second side 114 of the building may include an electrical connection between at least a second power source 124 and the power bus 130.

[0044] The power bus 130 may be a bus duct composed of superconducting cables. The superconducting cables may be configured to carry direct current (DC) power between about 10 kA and 30 kA. Due to the low resistance of the superconducting cables, the voltage at a given point of the superconducting cables may be between 1 kV and 40 kV.

[0045] Each superconducting cable may be connected to one or more server racks 140. The power bus 130 may be a power plane positioned directly above the server racks 140 to leave space at the data center floor. Alternatively, given the relatively small size of the superconducting cables compared to the conductive bus duct elements of a standard power bus, the power plane may be routed directly below the server racks 140, which may also leave space at the data center floor. As another alternative, the superconducting cables may be at floor level next to the server racks 140 without taking up excessive space.

[0046] The power bus 130 can be configured to transfer and guide power throughout the building through a network of busbar sections. Due to the low resistance in the superconducting cable, the power provided from the first side 112 of the building can be transferred to the second side of the building through the superconducting cable with minimal losses due to resistance. This means that high current density can be easily concentrated at any given location of the data center building, so the need for high current density can be met at any location. The ability to meet high current density needs includes even when two or more server racks in separate sections of the data center require high current density at the same time.

[0047] exist Figure 1 In the example of , the first side 112 and the second side 114 of the building are opposite sides of the building. However, in other examples, the first side and the second side can be adjacent sides of the building or non-opposite sides, as long as the power does not run through the building in a single downstream direction. Additionally, more than two power sources can be provided to the building. Figure 1 In the example of FIG. 4 , four power sources are shown, but in other examples, more or fewer power sources may be connected. It should also be understood that in other examples, power sources may be connected to the bus duct element from any side of the building. In addition, although Figure 1 A schematic layout of only one floor level of a data center is shown, but it should be understood that a data center may include multiple floor levels, each floor level including additional servers and server racks, which may have similar Figure 1 The same or similar layout of the floor layers shown in .

[0048] Figure 2 Figure 1 shows an example electrical connection between a superconducting cable power plane and a server rack in a data center. Figure 2In the system 200, a data center building 210 is connected to each of power source A 222 and power source B 224 from opposite sides of the building 210. Through the power connections at opposite ends of the building 210, the two power sources are connected to a superconducting cable network 230 inside the building. Figure 2 In the example of FIG. 2 , both power supplies 222, 224 are connected to the same superconducting cable so that power from both power supplies 222, 224 can be directed to any point within the superconducting cable network 230. Server racks 242, 244, 246 can be electrically connected to both power supplies 222, 224 through connections to the superconducting cables.

[0049] exist Figure 2 In the example of , the power plane 230 is shown as including an interface 250 for providing DC power to a DC bus duct 260. The DC bus duct 260 can carry power from the power plane to each of the connected server racks 1-N, 242, 244, 246. The power plane can be arranged in any of a variety of configurations, including but not limited to a mesh configuration, a ring configuration, and a serpentine configuration. Additionally, in some example arrangements herein, one or more controllers can be provided within the power plane to direct or distribute power through the power plane, which in turn can control the configuration of the power plane. Therefore, through the controller in the power plane, the power plane can be changed from a mesh configuration to a ring configuration or vice versa, or between any of the other example configurations.

[0050] The interface 250 can be designed to dock the superconducting cable of the power plane 230 to the copper plate of the copper conductor included in the DC bus duct 260. Superconducting cables are generally designed to operate at very low temperatures in order to maintain low resistance performance. In contrast, copper conductors are designed to conduct electricity at or near room temperature. Therefore, the interface 250 is specifically designed to provide an electrical connection between a very low temperature superconducting cable and a relatively hot bus duct, while maintaining the temperature difference and avoiding an increase in resistance or a decrease in efficiency at the superconducting cable. Additionally, the superconducting cable can have a high voltage of several thousand volts and a high current of several thousand amperes, while the DC bus duct can be configured to carry a medium voltage of several hundred volts, such as between 110 and 440 volts, and carry a relatively low current, such as a commercial grade current, such as about 20A. Therefore, the interface 250 is also specifically designed to provide an electrical connection between a high voltage line such as a superconducting cable and a medium voltage line such as a copper DC bus duct. Such interfaces have been developed in other applications with high efficiency and without degradation of superconducting cables, such as in the transmission of power for the Large Hadron Collider at CERN, and can be applied in a similar manner to distribute power to server racks in a data center. The DC bus duct 260 can be further connected to a plurality of server racks 1-N, 242, 244, 246 via one or more corresponding tap connectors 262, 264, 266.

[0051] Although for illustration purposes Figure 2 Only one DC bus duct 260 is shown in the figure, but it should be understood that the data center includes multiple DC bus ducts, each of which is connected to the power plane at various interface points distributed throughout the power plane. Each DC bus duct can have a relatively short length compared to the power plane and can be connected to a limited number of server racks. For example, the DC bus duct can have a length between about 10 meters and 100 meters and can be connected to hundreds of server racks. The use of short DC bus ducts provides a high degree of flexibility in configuring the data center, so that a wide range of power densities can be provided between different DC bus ducts and between different sections of the building.

[0052] The electrical connection between the power plane 230, the DC bus duct 260 and the server racks 242, 244, 246 can be controlled by one or more controllers (not shown). In some examples, the controller can be a manually operated physical switch accessible from within the data center building. The physical switches can be distributed throughout the building, concentrated in a specific location, or some combination of the two. In other examples, the controller can include one or more computing systems having one or more processors, a memory for storing instruction operations to be executed by the one or more processors, and an input / output component. The processor can be a well-known processor or other lesser-known type of processor, and can be a dedicated controller such as an ASIC. The memory can be a non-temporary computer-readable medium capable of storing information accessible by the processor, such as a hard drive, a solid-state drive, a tape drive, an optical storage device, a memory card, a ROM, a RAM, a DVD, a CD-ROM, a writable and read-only memory. The input / output component can be capable of sending and receiving data, such as receiving input signals such as user input or monitoring data about the data center, and sending control signals to control one or both of the interface 250 and the tap connectors 262, 264, 266.

[0053] Figure 3 Illustrations such as Figure 2 An example implementation of a power distribution system 300 using superconducting cables as in the system 200 of FIG. Figure 3 In the example of FIG. 1 , each server rack 342, 344, 346, 348 is connected to the server rack 342, 344, 346, 348 through the interface 350, the DC bus duct 360 and the like. Figure 2 The tap connection described above is connected to the power plane 330. The controller 310 is positioned along the DC bus duct 360 to control the opening and closing of the connection between the DC bus duct 360 and the server rack. Therefore, the controller is configured to provide control signals to the interface 350 and each tap connector through the DC bus duct 360, thereby controlling the power distribution from the power plane 330 to each server rack 342, 344, 346, 348. Figure 3 In an example of FIG. 4 , each DC bus conduit of a data center may be provided with a separate controller to control the distribution of power from a power source to server racks connected to the DC bus conduit.

[0054] Figure 4 Illustrations such as Figure 2 An alternative example implementation of a power distribution system 400 using superconducting cables as in the system 200 of FIG. Figure 4 In the example of FIG. 1 , each server rack 442, 444, 446, 448 is connected to the server rack 442, 444, 446, 448 through the interface 450, the DC bus duct 460 and the like. Figure 2The tap connection described is connected to the power plane 430. The corresponding controllers 412, 414, 416, 418 are located at the corresponding server racks 442, 444, 446, 448. Each controller is configured to control the opening and closing of the connection between the DC bus duct 460 and the server rack corresponding to the controller. Therefore, each controller is configured to manage the power distribution from each tap-connected DC bus duct 460 to the corresponding server rack, and then control the power distribution from the power plane 430 to each server rack 442, 444, 446, 448. Figure 4 In an example of , each server rack may be provided with a separate controller. Alternatively, a cluster of server racks may be provided with a single controller, thereby controlling power distribution on a cluster-by-cluster basis.

[0055] exist Figure 2-4 In the example of , each server rack is connected to two power supplies via a single superconducting cable connection. However, in other examples, the server racks may be connected to different power supplies via respective superconducting cable connections.

[0056] Figure 5 Another example electrical connection between a superconducting cable power plane and a server rack in a data center is shown. Figure 5 In the system 500, a data center building 510 is connected to each of a power source A 522 and a power source B 524 from opposite sides of the building 510. Power source A 522 is connected to a first power planar superconducting cable network 532 inside the building via a first power connection on a first side of the building 510. Power source B 524 is connected to a second power planar superconducting cable network 534 inside the building via a second power connection on an opposite second side of the building 510.

[0057] exist Figure 5 In the example of , the superconducting cable networks 532, 534 are not connected to each other, so that the power sources 522, 524 are connected to different superconducting cables. However, the two networks can be arranged to run in parallel with each other to provide power to any point within the building 510 from any one power source. In other example arrangements, the building can include connections for connecting the two networks to each other to form a single network. This may result in a superconducting cable network having at least some redundant cables.

[0058] Server racks 542, 544, 546 can be electrically connected to two power supplies 522, 524 from parallel power plane networks 532, 534 through connections to corresponding superconducting cables. Figure 5In the example of FIG. 5 , the first power plane 532 is shown as including an interface 552 for providing DC power to a first DC bus conduit 562. The first DC bus conduit 562 can carry power from the first power plane 532 to each of the connected server racks 1-N, 542, 544, 546. In parallel, the second power plane 534 is shown as including an interface 554 for providing DC power to a second DC bus conduit 564. The second DC bus conduit 564 can carry power from the second power plane 534 to each of the connected server racks 1-N, 542, 544, 546. Thus, each server rack can include multiple connections to connect to multiple power supplies.

[0059] Figure 6 Illustrations such as Figure 5 An example implementation of a power distribution system 600 using multiple superconducting cable networks connected to respective power sources as in the system 500 of FIG. Figure 6 In the example of FIG. 1 , each server rack 642, 644, 646, 648 is connected to the server rack 642, 644, 646, 648 through the first interface 652, the first DC bus duct 662 and the like. Figure 5 The tap connection described is connected to the first power plane 632. The corresponding controllers 612, 614, 616, 618 are located at the corresponding server racks 642, 644, 646, 648. Each controller is configured to control the opening and closing of the connection between the DC bus duct 662, 664 and the server rack corresponding to the controller. Therefore, each controller is configured to manage the power distribution from each tap connection of each DC bus duct 662, 664 to the corresponding server rack, and then control the power distribution from the two power planes 632, 634 to each server rack 642, 644, 646, 648. Figure 3 As with the example of , each server rack may be provided with a separate controller, or a cluster of server racks may be provided with a single controller, thereby controlling power distribution on a cluster-by-cluster basis.

[0060] In other examples, a controller may be provided instead of or in addition to Figure 6 The controllers 612, 614, 616, 618 shown in the example, for example, can be positioned along the first DC bus duct 662 to control the opening and closing of the connection between the first DC bus duct 662 and the server rack, and can be positioned along the second DC bus duct 664 to control the opening and closing of the connection between the second DC bus duct 664 and the server rack. This arrangement is similar to the arrangement described with respect to Figure 3 The arrangement described, but adapted to control power received from a plurality of power supplies via a single superconducting cable power plane network.

[0061] exist Figure 2-Figure 6 In the example of , each DC bus duct is connected to a corresponding superconducting cable. However, in other examples, the DC bus duct may include multiple superconducting cable connections in order to be connected to multiple power planes, or more particularly to multiple power supplies.

[0062] Figure 7 Another example electrical connection between a superconducting cable power plane and a server rack in a data center is shown. Figure 7 In the system 700, a data center building 710 is connected to each of a power source A 722 and a power source B 724 from opposite sides of the building 710. Power source A 722 is connected to a first power plane network of superconducting cables 732 inside the building via a first power connection on a first side of the building 710. Power source B 724 is connected to a second power plane network of superconducting cables 734 inside the building via a second power connection on an opposite second side of the building 710. Figure 6 As with the example system of , superconducting cable networks 732 , 734 are not connected to each other, but can be arranged to provide power to any point within building 710 from any one power source.

[0063] and Figure 5 As in the example of FIG. 7 , server racks 742, 744, 746 can be electrically connected to two power supplies 722, 724 from parallel power plane networks 732, 734 through connections to corresponding superconducting cables. Figure 7 In the example of FIG. 7 , both the first power plane 732 and the second power plane 734 are connected to a single DC bus conduit 760 to provide power to the server racks 742, 744, 746. Each power plane 632, 634 includes a respective interface 752, 754 for providing DC power to the single DC bus conduit 760. Thus, each server rack may include a single DC bus conduit connection to connect to a separate power plane.

[0064] Figure 8 Illustrations such as Figure 7 An example implementation of a power distribution system 800 using multiple superconducting cable networks connected to respective power sources as in the system 700 of FIG. Figure 8 In the example of FIG. 1 , each server rack 842, 844, 846, 848 is connected to both the first power plane 832 and the second power plane 834 through a single DC bus duct 860. The single DC bus duct 860 includes respective interfaces 852, 854 for connecting to the first power plane 832 and the second power plane 834. Respective controllers 812, 814 are positioned at the superconducting cables of the respective power planes 832, 834 to control whether power is provided to the DC bus duct 860 from each respective power source.

[0065] Fig. 9 Illustrations such as Figure 7 Another example implementation of a power distribution system 900 using multiple superconducting cable networks connected to respective power sources as in the system 700 of FIG. Fig. 9 In the example of , each server rack 942, 944, 946, 948 is connected to both the first power plane 932 and the second power plane 934 via a single DC bus duct 960. The single DC bus duct 960 includes respective interfaces 952, 954 for connecting to the first power plane 932 and the second power plane 934. A single controller 910 is positioned at the single DC bus duct 960 to control opening and closing of the connection between the DC bus duct 960 and the server racks 942, 944, 946, 948. Therefore, the controller 910 is configured to provide control signals to each respective interface 952, 954 and to the respective tap connectors of the server racks via the DC bus duct 960, thereby controlling the power distribution from each power plane 932, 934 to the respective server racks 942, 944, 946, 948. Fig. 9 In an example of a data center building, each DC bus conduit may be provided with a separate controller to control the distribution of power from the power source to the server racks connected to the DC bus conduit.

[0066] In other examples, a controller may be provided to supplement or replace Figure 8 and Fig. 9 For example, each server rack or each cluster of server racks may include its own controller for controlling the connection between the server rack and the DC bus conduit, with Figure 4 and Figure 6 The controller shown in the example arrangement is equivalent to .

[0067] It should be understood that Figure 3 , Figure 4 , Figure 6 , Figure 8 and Fig. 9 The example arrangements of are provided only as examples, and concepts from one example can be employed in other example arrangements. Figure 8 As shown in the example, Figure 3 , Figure 4 , Figure 6 and Fig. 9 The example arrangement of may be adapted to include a controller at a respective superconducting cable.

[0068] The above examples relate to power distribution elements with DC components, such as DC superconducting cables and DC bus ducts. However, it should be appreciated that the same concepts can be applied to alternating current (AC) components, such as AC superconducting cables and AC bus ducts. Thus, the example system can be arranged to deliver power to the server rack using either an AC power source or a DC power source.

[0069] Because the example system is connected to power sources on multiple sides of the building, there is no one side of the building through which all the current flows. Therefore, any particular side or component of the power bus cannot be characterized as "upstream" or "downstream" from another side or component. Certain components and sides may be "upstream" relative to a single power source, but may be "downstream" relative to different power sources. This provides several advantages over the typical arrangement with upstream elements, downstream elements, and unidirectional current flow. This can result in more uniform current density throughout the data center. Providing uniform current density can then avoid underutilization of distribution elements in the power plane, such as when the bus duct of a conventional power plane has a predefined capacity of power but is underutilized downstream where the current density is lowest. In addition, if power is cut off from one power source due to a failure of a distribution component on one side of the building, the server rack can still receive power from the power source connected to the other side of the building through the distribution element on the other side of the building.

[0070] The system described herein provides improved control of providing high power density to specific locations in a data center on demand. In conventional data centers where it is not possible to provide high current density to specific locations, it is often necessary to disperse individual server racks with high current density throughout the data center building space, rather than concentrating high-density server racks together in discrete locations such as aisles or quadrants of a building. For example, high current density may be required for server racks that include network equipment or server racks that perform machine learning operations such as in a machine learning data directory. In order to balance the power needs throughout the building, it may be necessary to disperse server racks with high current density because power distribution is limited in its configurability. However, server racks that must be dispersed with high current density require advance planning, and make it difficult to add more high current density or low current density server racks at a later time without disrupting the previous configuration. Planning in advance and disrupting server rack configurations increase the cost and time of maintenance operations. Additionally, some server racks with high current density may include confidential data that would be best separated in discrete corners or sections of the building rather than scattered throughout the building.

[0071] In contrast, the ability to provide high current density to specific locations without having to balance power density avoids the need to plan server rack configurations in advance. New high-density servers can be clustered together in a single area or section of a building, and the superconducting cables of the power plane can supply sufficient power to the area or section at a high current density without suffering electrical losses along the conductor elements. This approach can then allow server racks that require high current density and high confidentiality, such as MLDC server racks, to be concentrated in discrete parts of the building. Access to low-confidentiality servers can be granted, such as for maintenance operations, while access to high-confidentiality servers remains restricted.

[0072] Similarly, new low density servers can be clustered in a single area, and the shunt switches of the power bus can be controlled to make a power configuration that provides low current density in the location of the new server racks.

[0073] The above concepts apply equally to initially arranging high current density and low current density server racks within a data center when it is first established, as well as to subsequently adding new server racks after the data center has been established.

[0074] Example Method

[0075] Fig.10 1 is a flow chart of an example maintenance routine 1000 in accordance with the present application. The routine may allow for maintenance of separate sections of a data center at separate times. A first section of the data center may include a first portion of server racks, and a second section of the data center may include a second portion of server racks. The sections may be physically separated from one another so that one section may be disconnected from multiple power supplies without affecting the connection between the multiple power supplies and another section. This may be advantageous for performing maintenance on a portion of the server racks without affecting the operation of another portion of the server racks. It should be understood that the routines described herein are merely examples, and in other examples, certain steps may be added, subtracted, replaced, or reordered.

[0076] At box 1010, a first control signal may be provided from a controller for controlling the first portion of the server to electrically disconnect the first portion of the server from the connected power supply. The control signal may be provided automatically or manually. Additionally, the disconnection of the first portion of the server may be completed without disconnecting the other second portion of the server from the power supply. At box 1020, the disconnected first portion of the server may be maintained. At box 1030, a second control signal may be provided from a controller for controlling the first portion of the server to electrically reconnect the first portion of the server to the power supply. As with the disconnection process of box 1010, reconnecting the first portion of the server at box 1030 may be completed without disconnecting the second portion of the server from the power supply. Upon completion of boxes 1010-1130 of routine 1000, maintenance of the first portion of the server is completed.

[0077] In some examples, the second portion of servers may be maintained in the same manner as the first portion of servers. Boxes 1040-1160 describe maintenance of the second portion of servers. At box 1040, a third control signal may be provided from the controller for controlling the second portion of servers to electrically disconnect the second portion of servers from the connected power source. Disconnection of the second portion of servers may be accomplished without disconnecting the first portion of servers. At box 1050, maintenance may be performed on the disconnected second portion of servers. At box 1060, a fourth control signal may be provided from the controller for controlling the second portion of servers to electrically reconnect the second portion of servers to the power source, again without disconnecting the first portion of servers from the power source in the process.

[0078] At any time during the maintenance routine 1000, disconnection of an "upstream" element does not cause disconnection of a "downstream" element. This is because the system of the present disclosure provides multi-directional current flow and lacks elements that can be absolutely characterized as "upstream" and "downstream."

[0079] Additionally, maintenance routine 1000 allows for separate maintenance of high confidentiality servers and low confidentiality servers based on their location because, in the example system described above, servers can be clustered based on properties such as confidentiality without having to worry about load balancing or advance layout planning. As described above, conventional data centers typically require high current density server racks to be dispersed within the physical space of the data center building in order to provide proper load balancing between the building's power distribution elements. In contrast, the example arrangements of the present disclosure allow high current density server racks to be concentrated in discrete physical locations within the data center building. By extension, server racks with high confidentiality levels, such as MLDCs or network equipment, can be concentrated in more secure sections of the building, thereby providing access to lower confidentiality server racks while maintaining restricted access to higher confidentiality server racks.

[0080] Although the technology herein has been described with reference to specific embodiments, it should be understood that these embodiments are merely illustrative of the principles and applications of the technology. Therefore, it should be understood that various modifications may be made to the illustrative embodiments and other arrangements may be designed without departing from the spirit and scope of the technology as defined by the appended claims.

[0081] Most of the above-mentioned alternative examples are not mutually exclusive, but can be implemented in various combinations to achieve unique advantages. Because these and other variations and combinations of the features discussed above can be utilized without departing from the subject matter defined by the claims, the above description of the embodiments should be performed by illustration rather than by limiting the subject matter defined by the claims. As an example, the previous operations do not have to be performed in the above-mentioned precise order. On the contrary, the various steps can be processed in different orders (such as oppositely or simultaneously). Unless otherwise stated, steps can also be omitted. In addition, the provision of the examples described herein and the clauses represented by phrases such as "such as", "including", etc. should not be interpreted as limiting the subject matter of the claims to specific examples; on the contrary, these examples are intended to illustrate only one of many possible embodiments. In addition, the same reference numerals in different drawings can identify the same or similar elements.

Claims

1. A system for providing electric power to a building, the system comprising: a superconducting cable network configured to supply power to a plurality of server racks arranged within the space of the building; a first power connection configured to connect a first power source from a first side of the building to the superconducting cable network and configured to supply power to the superconducting cable network; a second power connection configured to connect a second power source from a second side of the building different from the first side to the superconducting cable network and configured to supply power to the superconducting cable network; as well as a plurality of bus ducts, each bus duct being configured to connect the superconducting cable network to one or more of the plurality of server racks, wherein the first power connection and the second power connection are configured to collectively supply power to the superconducting cable network so as to provide power from any one power source to any point within the building, and are arranged such that power supplied from a first side of the building passes through the building in a first downstream direction and such that power supplied from a second side of the building passes through the building in a second downstream direction, wherein the second downstream direction is different from the first downstream direction.

2. The system according to claim 1, wherein: The first side and the second side are opposite sides of the building.

3. The system of claim 1, further comprising a plurality of switch controllers, each switch controller being configured to control an electrical connection between at least one of the first power connection and the second power connection and at least one server rack.

4. The system according to claim 3, wherein: The plurality of switch controllers include a superconducting switch positioned between a first section and a second section of the superconducting cable network and configured to control an electrical connection between the first section and the second section of the superconducting cable network.

5. The system according to claim 3, wherein: The plurality of switch controllers include an electrical switch positioned at a first bus duct and configured to control an electrical connection between the first bus duct and one or more server racks connected to the first bus duct.

6. The system according to claim 3, wherein: The plurality of switch controllers include an electrical switch positioned at a first server rack and configured to control electrical connections between a plurality of bus ducts connected to the first server rack and the first server rack.

7. The system according to claim 3, wherein: The plurality of server racks are divided into discrete sections, wherein each discrete section is connected to each of the first power supply and the second power supply, and wherein the plurality of switch controllers are configurable to disconnect any one of the discrete sections of the plurality of server racks from the first power supply and the second power supply while other sections of the plurality of server racks remain connected to at least one of the first power supply and the second power supply.

8. The system according to claim 1, wherein: The network of superconducting cables is arranged to allow electricity to flow therethrough in a serpentine configuration.

9. The system according to claim 1, wherein: The network of superconducting cables is arranged to allow power to flow therethrough in a toroidal configuration.

10. The system according to claim 1, wherein: The network of superconducting cables is arranged to allow power to flow therethrough in a mesh configuration.

11. The system according to claim 1, wherein: The superconducting cable network is positioned directly above the plurality of server racks.

12. The system of claim 1, wherein: The space of the building includes: a first section, the first section including a first part of servers assigned with a relatively high priority; and a second section, the first section including a second part of servers assigned with a relatively low priority.

13. The system according to claim 12, wherein: The first part of servers includes at least one of a machine learning server and a network device.

14. The system of claim 1, wherein: The superconducting cable network is configured to transmit a direct current (DC) voltage between 1 kV and 20 kV.

15. The system of claim 14, wherein: The superconducting cable network is configured to transmit a DC current between 10 kA and 20 kA.

16. A data center, comprising: The system according to claim 1; and The plurality of server racks are arranged within a space in a building.

17. A server maintenance method for updating a portion of servers in a data center, the data center comprising a superconducting cable network and one or more controllers, the superconducting cable network being configured to supply power from each of a first power source and a second power source to a plurality of server racks arranged in a space of a building, wherein: The superconducting cable network is connected to the first power source via a first power connection positioned on a first side of the building and to the second power source via a second power connection positioned on a second side of the building different from the first side, wherein the first power connection and the second power connection are configured to collectively supply power to the superconducting cable network so as to provide power from any one power source to any point within the building, and are arranged such that power supplied from the first side of the building traverses the building in a first downstream direction and such that power supplied from the second side of the building traverses the building in a second downstream direction, wherein the second downstream direction is different from the first downstream direction, and the one or more controllers are configured to control electrical connections between the first power source and the second power source and the plurality of server racks, the method comprising: providing a first control signal from the one or more controllers, the first control signal being used to electrically disconnect a first portion of servers from the first power source and the second power source without disconnecting a second portion of servers from the first power source and the second power source; Maintaining the first part of servers; providing a second control signal from the one or more controllers, the second control signal being used to electrically reconnect the first portion of servers to the first power source and the second power source without disconnecting the second portion of servers from the first power source and the second power source; providing a third control signal from the one or more controllers, the third control signal being used to electrically disconnect the second portion of servers from the first power source and the second power source without disconnecting the first portion of servers from the first power source and the second power source; Maintaining the second part of servers; and providing a fourth control signal from the one or more controllers, the fourth control signal being used to electrically reconnect the second portion of servers to the first power source and the second power source without disconnecting the first portion of servers from the first power source and the second power source, The first part of servers and the second part of servers are divided according to corresponding confidentiality levels.

18. The server maintenance method according to claim 17, wherein: The first control signal electrically disconnects a first section of the superconducting cable network from a second section of the superconducting cable network, and the second control signal electrically reconnects the first section of the superconducting cable network to the second section of the superconducting cable network.

19. The server maintenance method according to claim 17, wherein: The first control signal electrically disconnects the first portion of servers from the superconducting cable network at one or more DC bus conduits configured to provide power from the superconducting cable network to the first portion of servers, and wherein the second control signal electrically reconnects the first portion of servers to the superconducting cable network at the one or more DC bus conduits.

20. The server maintenance method according to claim 17, wherein: The first control signal electrically disconnects a first server rack of the first portion of servers from each of a first DC bus conduit and a second DC bus conduit, wherein the first DC bus conduit is configured to provide power to the first server rack from the first power supply and the second DC bus conduit is configured to provide power to the first server rack from the second power supply, and wherein the second control signal electrically reconnects the first server rack to the first DC bus conduit and the second DC bus conduit.

Images