A two-dimensional torus topology is an interconnection topology used in high performance computing applications. The topology interconnects nodes in a mesh-like fashion where the nodes are connected to their left, right, above, and below neighbors with the end nodes looped around so each node is connected to four neighbors.
Non-limiting examples of the present disclosure are described in the following description, read with reference to the figures attached hereto and do not limit the scope of the claims. In the figures, identical and similar structures, elements or parts thereof that appear in more than one figure are generally labeled with the same or similar references in the figures in which they appear. Dimensions of components and features illustrated in the figures are chosen primarily for convenience and clarity of presentation and are not necessarily to scale. Referring to the attached figures:
In the following detailed description, reference is made to the accompanying drawings which form a part hereof, and in which is depicted by way of illustration specific examples in which the present disclosure may be practiced. It is to be understood that other examples may be utilized and structural or logical changes may be made without departing from the scope of the present disclosure.
To form a torus from a mesh topology, large systems using the two-dimensional torus topology require long wires between the end nodes or the nodes on the edge of the arrays. Completing the torus using long wires introduces signal integrity problems due to long trace length between the two end nodes. Adaptations to the two-dimensional torus topology include connecting every other node and bringing some of the nodes to the back of the topology to make the wires shorter in length. Connecting every other node doubles the connections and wire lengths which results in a drastic increase in the printed circuit board layers and increases the cost.
Bringing some of the nodes to the back of the topology changes the configuration of the input and output ports. For example, the change in the configuration connects two right facing ports to one another and/or two left facing ports to one another in some instances of the nodes while keeping the other nodes connected with one left facing port and one right facing port. The mapping of the ports is inconsistent when the configuration allows for the mating of left facing ports with other left facing ports and/or mating of right facing ports to other right facing ports in some instances and mating right facing ports and left facing ports in other instances. Moreover, such inconsistencies make it difficult to manage a mapping of nodes. Additionally, the topology adaptations connecting every other node and bringing some of the nodes to the back of the topology are not compatible with a system that includes nodes of varying widths and/or lengths that are connected to one another in a single system.
In examples, a printed circuit board with a two-dimensional torus topology is provided. The printed circuit board includes an array of board connectors, an input port and an output port, and a set of interconnections. Each board connector of the array of board connectors is formed to receive an electronic connector The input port and the output port are formed on each board connector to provide an identical trace routing direction. The set of interconnections are formed to connect each board connector to another board connector. The signal path between each of the interconnections has a length less than a distance between a first board connector and a last board connector in the array of board connectors.
As used herein, the phrase “electronic module” refers to a computing device such as a server, a blade server, or a server cartridge that provides computer solutions, storage solutions, network solutions and/or cloud services.
As used herein, the phrase “compute node” refers to data circuit terminating equipment (DCE), such as a modem, hub, bridge or switch; data terminal equipment (DTE); or a distributed system node, such as a client, a server or a peer device.
As used herein, the term “interconnection” refers to connection via for example a wire.
As used herein, the phrase “two-dimensional torus topology” refers to a network of nodes where each node is connected to more than one of the other nodes and all nodes are connected by a routing technique.
As used herein, the term “input port” refers to a connection point on a printed circuit board for an electronic device, such as an electronic module or compute node. The input port route the flow of interconnections into the port or connection on the printed circuit board. The input port may include, for example, a plurality of input pins, such as gold pins to form an electrical connection between the printed circuit board and the electronic device.
As used herein, the term “output port” refers to a connection point on a printed circuit board for an electronic device, such as an electronic module or compute node. The output port routes the flow of interconnections out of the port or connection on the printed circuit board. The output port may include, for example, a plurality of output pins, such as gold pins to form an electrical connection between the printed circuit board and the electronic device.
As used herein, the phrase “signal path” refers to the interconnection in a route that a signal travels between board connectors and slots.
As used herein, the phrase “shortened signal path” refers to a reduced or shorter route for signals to travel between board connectors and slots. For example, a length of an interconnection is shortened to reduce or shorten the signal path.
As used herein, the phrase “trace routing direction” refers to a direction of an interconnection between two slots. For example, a direction of the interconnection between an output port and the following input port on a printed circuit board.
As used herein, the phrases “identical trace routing direction,” “same trace routing direction,” and “consistent trace routing direction” refer to a symmetry between at least two trace routing directions. For example, each input port and each output port is located in the same position on all board connectors, slots, or nodes, such as all board connectors, slots, or nodes having the input ports and the output ports in the same position or in a symmetrical layout compared to one another.
As used herein, the phrase “physically adjacent” refers to interconnected positions physically adjacent or neighboring one another in physical space.
The set of interconnections 140 are formed to connect each board connector to another board connector in the array of board connectors 120. The set of interconnections 140 may include, for example, copper traces. Each interconnection of the set of interconnections 140 include a shortened signal path between a first board connector and a last board connector, the shortened signal path includes a length less than a distance between the first board connector and the last board connector in the array of board connectors 120, the first board connector and the last board connector are arranged at opposite ends of the printed circuit board 100.
The two-dimensional torus topology 200 illustrated also provides an identical trace routing direction 230 for each board connector 220. The input ports 122 are positioned at a first position P1 on each board connector 220, and the output ports 124 are positioned at a second position P2 on each board connector 220 to provide the identical trace routing direction 230 between the set of interconnections 140.
The input ports 122 to connect to an input electronic connector of the electronic module, and the output ports 124 to connect to an output electronic connector of the electronic module. The identical trace routing direction 230 enables the electronic modules to be interchangeably inserted into each board connector 220. For example, each board connector 120 may interchangeably receive electronic modules. The electronic module includes at least one module size selected from a single length and single width module, a double width module, a double length module, a triple length module, and a triple width module. The electronic modules may also include at least one module type selected from a network switch, a storage module, and a compute module interchangeably.
Each electronic module 360 is coupled to another electronic module 360 via the set of interconnections 140. Each electronic module has at least one input connection 362 and at least one output connection 364. The at least one input connection 362 is formed to connect to the corresponding input port 122 and the at least one output connection 364 is formed to connect to the corresponding output port 124.
The set of interconnections 140 that form along the same trace routing direction 230 include a shortened signal path. The length of at least one of the interconnections is reduced or shortened by directly connecting the first board connector, labeled slot 1, to another board connector between the second board connector, labeled slot 2, and the second to last board connector, labeled slot 14. Similarly, the last board connector, labeled slot 15, is directly connected to another board connector between the second board connector, labeled slot 2, and the second to last board connector, labeled slot 14. Using such an arrangement, at least one interconnection 240 of the set of interconnections 140 is not directly connected to the slot in a position physically adjacent thereto and the arrangement reduces or shortens a length of the set of interconnections 140 therebetween. The interconnections between adjacent slots, or the interior interconnections 450, may remain unchanged in most instances. The end slots or first slot, labeled slot 1, and last slot, labeled slot 15; however, will have a signal path with a length that is reduced or shorter, i.e., a reduced or shortened signal path 142. In other words, the reduced or shortened signal path 142 is a percentage less than one-hundred percent of a distance between a first board connector associated with a first slot, labeled slot 1, and a last board connector associated with a last slot, labeled slot 15.
For example, the shortened signal path 142 is illustrated as follows. The signal path is formed between the first slot, labeled slot 1, that is connected to an interior slot, labeled slot 9, via a modified input interconnection 432. The modified input interconnection 432 is eight slot positions away instead of fifteen slot positions if the first slot, labeled slot 1, and the last slot, labeled slot 15, were connected as in typical topologies. Similarly, the last slot, labeled slot 15, is connected to an interior slot, labeled slot 7, via a modified output interconnection 434, which is eight slot positions away. The output of the slot, labeled slot 7, and the input of the slot, labeled slot 9, will remain the same. The modification to the first slot, labeled slot 1, and the last slot, labeled slot 15, form the shortened signal path 142 that approximately sixty percent of the distance between the first slot, labeled slot 1, and the last slot, labeled slot 15. The reduction in the signal path between the first slot, labeled slot 1, and the last slot, labeled slot 15, also requires some adjustments to the interconnections 140 to the output port 124 of the slot labeled 6 and the interconnections 140 to the input port 122 of the slot labeled 10, which are interconnected via a modified interior interconnection 436.
Moreover, the two-dimensional torus topology 200 also preserves the trace routing direction 230 between the input ports 122 and the output ports 124, which enable each slot 400 to be compatible with a variety of electronic modules 360. For example, the electronic modules 360 may include at least one module size selected from a single length and single width module 410, a double length module 420, a triple length module 430, a double width module 440, a triple width module 460, and a double length and double width module 480. The variety in the size of the electronic modules 360 also provides the ability to select different types of electronic modules 360, such as a network switch, a memory module, a storage module, and a compute module. Therefore, a single printed circuit board 100 may receive one or a combination of the sizes and types of electronic modules 360 as illustrated in
Referring to
Although the flow diagrams 600, 700 of
The present disclosure has been described using non-limiting detailed descriptions of examples thereof and is not intended to limit the scope of the present disclosure. It should be understood that features and/or operations described with respect to one example may be used with other examples and that not all examples of the present disclosure have all of the features and/or operations illustrated in a particular figure or described with respect to one of the examples. Variations of examples described will occur to persons of the art. Furthermore, the terms “comprise,” “include,” “have” and their conjugates, shall mean, when used in the present disclosure and/or claims, “including but not necessarily limited to.”
It is noted that some of the above described examples may include structure, acts or details of structures and acts that may not be essential to the present disclosure and are intended to be exemplary. Structure and acts described herein are replaceable by equivalents, which perform the same function, even if the structure or acts are different, as known in the art. Therefore, the scope of the present disclosure is limited only by the elements and limitations as used in the claims.
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/US2013/052766 | 7/30/2013 | WO | 00 |