CN108476080B - Multidimensional optoelectronic switch - Google Patents

Abstract

For transmitting optical signals from input means to output meansLA optoelectronic switch, the optoelectronic switch comprising: multiple leaf switches, each leaf switch having a radixRAnd are prepared byLDimension array arrangement, wherein each dimensioniHave corresponding sizeR _i (i=1、2、...、L) Each leaf switch having an association with coordinatesLTuple (x) ₁ 、...、x _L ) Giving it a relation toLA location of each of the dimensions; wherein each leaf switch isLMembers of a sub-array, theLEach of the sub-arrays and theLDimension (Wei)Different dimensions in degrees are associated and include: a plurality ofR _i Leaf switches with co-ordinates only in the second placeiDimensionally differentiated, each leaf switch having means for connecting to an input or output deviceCIndividual client ports and for connection to a spine switchFAn architecture port; a plurality ofS _i Spine switches each having the fabric ports for connecting to the leaf switchesRA plurality of fabric ports, and wherein in a given sub-array, each leaf switch in the sub-array is connected to each spine switch via an optically active switch.

Description

Multidimensional optoelectronic switch

FIELD

One or more aspects of embodiments according to the present invention relate to optoelectronic switches, and more particularly, to multi-dimensional optoelectronic switches and switching elements used therein.

Background

The current and continuing growth in data traffic and the data center's requirements for switching speeds and reduced energy consumption have driven a number of recent innovations. In particular, it has been recognized that optical switching provides many desirable attributes but that optical devices need to be controlled by and connected to electronic devices including conventional electronic data servers.

The optical device itself does not necessarily reduce the size or complexity of the switch. In order to improve the assembly and application flexibility of the optical switching unit, it is desirable to improve the scalability of the optical switch. One way to do this involves the topology of the components within the switching network. It is desirable to produce highly scalable optical switching units. Therefore, there is still a need for packet switching that optimally benefits from the speed of optics and the flexibility of CMOS electronics assembled in architectures suitable for great scalability.

SUMMARY

At a higher level, the general architecture of the present invention involves an array of leaf switches (which are client side switches to which input devices and output devices can be connected) that includes a number of sub-arrays, each sub-array being associated with a different dimension, and where switching can occur within a given sub-array along the "direction" of that dimension via a spine switch (which is a fabric side switch, with no external connections). To most easily understand how signals can be switched from a source leaf switch to a destination leaf switch, consider a two-dimensional array, where the source leaf switch and destination leaf switch are positioned in different rows (which are sub-arrays associated with a "horizontal" dimension) and columns (which are sub-arrays associated with a "vertical" dimension). The signals will then be exchanged in two hops: the first hop causes the signal from the source leaf switch to reach another intermediate leaf switch in the same row, which is in the same column as the destination leaf switch. Subsequently, in a second hop, the signal is switched along the column from the intermediate leaf switch to the destination leaf switch. Thus, it can be seen that when an optoelectronic switch extends to L dimensions, a signal (which may be a packet, for example) can be switched from any source leaf switch to any destination leaf switch in up to L hops, each hop being limited to a single sub-array.

An optoelectronic switch such as this is desirable to satisfy as many external devices as possible. This can be achieved in two main ways. The first way is to increase the number of dimensions in the array and the second way is to increase the size of the dimensionsR _i . Increasing the number of dimensions necessarily requires the inclusion of larger leaf switches in order to accommodate switching in more than one dimension. Increased dimension ofR _i It may mean that a spine switch used to control switching from, for example, a source leaf switch to an intermediate leaf switch within a given dimension needs to have a large radix (radix) in order to accommodate switching between all leaf switches in the sub-array, preferably in a non-blocking manner and with a reservation of bisection bandwidth. In the present invention, additional switching elements are included in each sub-array, and more particularly, a first aspect of the present invention provides for transmitting optical signals from an input device to an output deviceLA optoelectronic switch, the optoelectronic switch comprising:

multiple leaf switches, each leaf switch having a radixRAnd are prepared byLDimension array arrangement, wherein each dimensioniHave corresponding sizeR _i (i = 1、2、...、L) Each leaf switch having an association with coordinatesLTuple (x ₁ 、...、x _L ) Giving it a relation toLA location of each of the dimensions;

wherein each leaf switch isLMembers of a sub-array, theLEach of the sub-arrays and theLDifferent ones of the dimensions are associated and include:

a plurality ofR _i Leaf switches with co-ordinates only in the second placeiDimensionally differentiated, each leaf switch having means for connecting to an input or output deviceCIndividual client ports and for connection to a spine switchFAn architecture port;

a plurality ofS _i Spine switches each having fabric ports for connection to leaf switchesRA plurality of fabric ports, and

wherein in a given sub-array, each leaf switch in the sub-array is connected to each spine switch via an optically active switch.

Inserting an optically active switch in a leaf switchBetween spine switches means that fewer spine switches can be used to provide switching between the same number of leaf switches. Thus, in a preferred embodiment of the present invention, the number of leaf switches is greater than the number of spine switches in a given sub-array. In a given sub-array, there may be more leaf switches than the number of fabric ports per spine switch. This can be derived from considering cardinalityRTo understand that the cardinality is the same for both leaf switches and spine switches. In which L is>1, the leaf switch requires at least one fabric port per dimension for switching, and thus, there are fewer fabric ports available for switching within any given sub-array. In a sub-array, a fabric port on a spine switch involves switching only in the dimension associated with the sub-array. Thus, in a balanced network, fewer spine switches can support more leaf switches because of the multi-dimensional nature of optoelectronic switches. In a preferred embodiment of the process according to the invention,Lequal to two or more and may take any of the following values: 2. 3, 4, 5, 6, 7, 8, 9, 10.

In embodiments of the present invention, switching is provided between leaf switches within a given sub-array by a spine switch, which may be an electronic packet switch and may be referred to as an EPS or AOBM. It should also be noted that throughout this application, a leaf switch may be referred to as an "optical packet processing module" or "OPPM".

To understand the advantageous effects of the present invention, consider a sub-array in which spine switches are connected only via a single spine switch or a group of parallel spine switches ("parallel" meaning that a spine switch is connected only to leaf switches and not to other spine switches). Each of these spine switches must necessarily be connected to each leaf switch in the sub-array in question, and each leaf switch must be connected to each spine switch, otherwise it may not be possible to provide a connection between each leaf switch and all other leaf switches in the sub-array.

Thus, as mentioned above, for larger sub-arrays, a large number of large radix spine switches are required. Each leaf switch still has to be connected to each spine switch, but in the present invention the leaf switches are connected to the spine switches via optically active switches. In this context, an "active switch" is a switch that: where the path of signals through the switch (e.g., from its inputs to its outputs) can be actively controlled and changed to provide virtually full network capability without requiring full network links within the switch. The optically active switch may be an optical circuit switch or an OCS (the terms are used interchangeably herein). More specifically, each optically active switch may have a number of inputs and a number of outputs, the number of inputs being the same as the number of outputs. The "radix" of an optically active switch refers to the number of inputs or outputs, not the total number of inputs and outputs. Thus, the output of a given optically active switch may be connected to multiple or a group of spine switches in a sub-array. A given leaf switch may be connected to a cluster of optically active switches, each of which is connected to a set of spine switches. More specifically, each of the clusters of optically active switches to which a given leaf switch is connected may be connected to a different set of disjoint spine switches ("disjoint" meaning that two optical switches in a given cluster are not connected to the same spine switch, i.e., there is no overlap of spine switch groups).

In this way, the leaf switches in question are connected to all spine switches in the sub-array, but via the smaller radix optically active switches. Since the optically active switch also has multiple inputs, these inputs can be shared among multiple leaf switches. In other words, for a given cluster of optical switches (whose outputs provide connections to all the spine switches in the sub-array), each input of each optically active switch in the cluster may be connected to a corresponding (i.e., different) leaf switch. Thus, leaf switches may be divided into multiple clusters, and in particular: in a given sub-array, the leaf switches may be divided into a plurality of clusters, each containing a plurality of leaf switches. Each cluster may itself have an associated cluster of optical active switches via which the optical active switches provide connections between each of the leaf switch clusters in the array and each of the spine switches. In other words, each cluster of leaf switches may be associated with a cluster of one or more fabric ingress optical active switches to form a line card assembly, each leaf switch in the cluster may be connected to each fabric ingress optical active switch in the line card assembly, and the fabric ingress optical switches may be arranged such that optical signals pass through the fabric ingress optical active switches during transmission from the leaf switches to the spine switches.

Optionally, in a given sub-array, each spine switch may be connected to a fabric ingress optical active switch in a line card assembly positioned in the sub-array, and no more than one fabric ingress optical active switch in the line card assembly is connected to the given spine switch. In the line card assembly, there may beLGroups of different fabric ingress optically active switches, each group configured to transmit optical signals within a respective sub-array containing the line card assemblies, each of the sub-arrays associated withLRespective ones of the dimensions are associated.

In order to maximize topological regularity, it is preferred that all or substantially all of the optically active switches employed in embodiments of the present invention have the same radix, or more specifically, the same number of inputs and outputs. In particular, when all optically active switches in a given sub-array have a given radix, it is possible for the same number of spine switches to support a number of leaf switches increased by a factor of the radix without compromising split bandwidth. For example, if a radix-3 optically active switch is used, the sub-array may be three times the size. In other words, the opportunity to scale up the array is greatly increased without increasing the size of the spine switch employed.

The above-described connections involve the transfer of signals from the output of the leaf switch to the input of the spine switch. To complete the transmission of a signal from one leaf switch in a sub-array to another leaf switch in the sub-array, the signal must be sent from the output of the spine switch to a fabric port on the destination (or intermediate) leaf switch. Thus, in some embodiments, such transmission is also via an optically active switch. Thus, each leaf switch in the sub-array may be connected to a second or "fabric egress" cluster of optically active switches. For clarity, the arrangement described in the previous paragraph is henceforth referred to as a "fabric ingress" cluster of optically active switches. Briefly, the fabric egress cluster has a mirrored configuration of the fabric ingress cluster. In other words, the input of a given optical active switch in the fabric egress cluster is connected to the output of a set of spine switches accordingly, and the input of each optical active switch in the cluster is connected to the output of a different set of disjoint spine switches. Each output of a given optical active switch in a fabric egress cluster is then connected to a fabric port of each leaf switch in the leaf switch cluster associated with the fabric egress cluster, and thus each fabric port on a given leaf switch in a cluster is connected to a different optical active switch within the fabric egress cluster associated with the leaf switch cluster. Other leaf switch clusters may be defined, for example, there may be a cluster containing one leaf switch from each fabric ingress cluster.

Each cluster of leaf switches may be further associated with one or more clusters of fabric egress optical active switches to form a fabric egress arrangement, each leaf switch in a given cluster may be connected to each fabric egress optical active switch in the fabric egress arrangement, and the fabric egress optical active switches may be arranged such that optical signals pass through the fabric egress optical active switches during transmission from the spine switch to the leaf switches. Each spine switch may be connected to a fabric egress optically active switch positioned in a fabric egress arrangement in the sub-array, and no more than one fabric egress optically active switch in the fabric egress arrangement is connected to a given spine switch.

When the leaf switches, optically active switches and spine switches are as described above, within a given sub-array, the leaf switches are effectively connected via a five-stage Clos network, i.e., leaf switch → (fabric ingress) optically active switchingMachine → spine switch → (fabric egress) optically active switch → leaf switch. As will be apparent from the above description, the fabric ingress and egress optical active switches are different entities, but there is only one set of leaf switches. This network is therefore best represented (as will be seen in more detail later in this application) as an unfolded Clos network, but strictly speaking it can be considered as a partially folded Clos network. The use of a Clos or partially folded Clos network such as this means that it is possible to adapt in a non-blocking manner to the sub-arraysR _i Any combination of one-to-one pairings between leaf switches.

The arrangement described in the preceding paragraph need not be used for swapping in all L dimensions. In some embodiments of the invention, with respect only to the sub-arrays associated with the M dimensions, each leaf switch in a given sub-array may be connected to each spine switch via an optically active switch, where M < L. The other optional features presented above may be applied to any or all dimensions in which switching between leaf switches is done via optically active switches.

The optoelectronic switch according to the first aspect of the invention may be arranged on an optical backplane. Thus, the component parts (i.e., leaf switches, optically active switches, and spine switches) may be positioned on the card. The card may be a printed circuit board on which electronic components, optical components, and control components (i.e., an arbiter) are formed. The card may also be accommodated between both the optical component and the electronic component. In particular, embodiments of the present invention may include two types of cards: line cards and fabric cards. More specifically, the components may be located on a line card or fabric card. Line cards are "client-oriented" cards, while fabric cards are "fabric-oriented" cards.

The leaf switches and fabric ingress optically active switches (i.e., line card assemblies) may be located on respective line cards. More specifically, a single line card includes at least one leaf switch and at least one fabric ingress optically active switch. In some embodiments, a single line card may include multiple leaf switches and/or multiple fabric ingress optically active switches. In embodiments where there is a leaf switch cluster and an associated fabric ingress optical active switch cluster, the leaf switch cluster and its associated fabric ingress optical active switch cluster are preferably installed on the same line card. In the N-dimensional embodiment of the invention, each leaf switch is a member of N sub-arrays, as previously discussed in this application.

In embodiments where there are leaf switch clusters and fabric ingress optical active switch clusters, each leaf switch cluster and its associated fabric ingress optical active switch cluster may be located on their own line cards. In a given sub-array, the spine switch and fabric egress optical switch may be located on a fabric card. The fabric card may also include an arbiter for controlling the path of signals through a fabric egress optical active switch located on the fabric card. The fabric card may also include a plurality of arbiters, each configured to control the path of signals through a respective fabric egress optical active switch on the fabric card. Not all ridges need be on the same fabric card. However, it is advantageous to package them such that the fabric-egress OCS connected to the same spine are all packaged on the same card as the spine. In practice, this means that the spine and the cluster of fabric-egress OCS are packaged on the same card (in a similar manner to the leaf and fabric-ingress OCS).

There are two control elements in the optoelectronic switch of the present invention: route/load balancing, and arbitration.

The packet processor makes routing decisions based on the packet's destination address and current location. On the path from the leaf to the spine, a routing decision selects which spine to route to (typically attempting to balance the load on the available spines), and then determines the specific output port on both the local leaf switch and the fabric ingress optically active switch. The output identifier of the fabric ingress optically active switch is passed to the arbiter so that it can determine which of its inputs need to be connected to which of its outputs. On the path from the spine to the leaf, the routing decision selects the appropriate leaf based on the destination of the packet, which in turn determines the local output ports of the spine and fabric egress optical active switches.

Arbitration is performed by an arbiter and is a process whereby it is determined which path the signal should take to traverse the optically active switch, i.e. from which input to which output, in order to ensure that all signals incident on the optically active switch are directed to the correct next switching element (which may be a spine switch or a leaf switch, depending on which "stage" the signal is at). Thus, in some embodiments, there may be an arbiter associated with each optically active switch in a given sub-array, in other words, a line card may include an arbiter for controlling the path of signals through fabric ingress optically active switches included in a line card assembly located on the line card, or there may be multiple arbiters, each configured to control the path of signals through a respective fabric ingress optically active switch. Alternatively, because the optically active switch may have a small radix (e.g., 2, 3, 4, 5, 6, 7, or 8), the arbitration process is relatively simple (e.g., compared to a radix 24 switching element), and multiple arbiters may be combined into a single arbitration component, which may be an ASIC. In some embodiments, there may be a single arbiter, or a single arbitration component as described above located on each line card. In the embodiments described in this paragraph, the control performed by the arbiter is limited to the scope of the line card in question. This minimizes the delay and synchronization issues associated with the control plane: the distance/time of flight on the card can be controlled to tighter tolerances within the physical dimensions of individual cards that can be positioned at a substantial distance from each other on the card. Furthermore, by having several arbiters, each associated with a smaller number (e.g., one) of optically active switches, a large number of small problems can be solved quickly and in parallel, as opposed to having to focus on solving more complex problems relatively slowly.

Broadly, the optically active switches are controlled by an arbiter configured to control the action of at least one of the optically active switches and the spine switch within a given sub-array based on destination information stored in the data packets to be switched. This then allows routes to be provided which ensure that all data reaches the appropriate leaf switch in a non-blocking manner and which minimize the occurrence of bottlenecks. The packet processors in the leaf switches may each be connected to an arbiter. When a data packet is received at a transmit-side packet processor, it may send a request to an arbiter to which the packet processor is connected, the request preferably identifying the destination leaf switch or identifying the next leaf switch (which may be the destination leaf switch) to which the data packet should be sent. The arbiter can then establish a scheme that ensures, to the greatest extent possible, that each packet can perform its next hop. The structure of the leaf switches is described in more detail below.

The arbiter may be connected to other components, such as packet processors and optically active switches, using dedicated control channels. They may also be connected to a driver chip configured to control the action of the optically active switch.

A second aspect of the invention provides a leaf switch that may be used in an optoelectronic switch according to the first aspect of the invention. In particular, the leaf switch may include:

a client port for connection to an input device or an output device and a client portion connected to the client port;

a first fabric portion and a second fabric portion connected to a first fabric port and a second fabric port, respectively, to process signals and communicate with a spine switch or an optically active switch, the first fabric portion having a transmit side and a receive side,

wherein the transmitting side comprises:

a transmit-side input to receive a first electronic signal carrying information, the information comprising information about a destination leaf switch of the first electronic signal, the first electronic signal received from any one of:

an output of the second architecture part, or

An input device via the client portion and the client port;

a modulator for converting the first electronic signal into a first plurality of optical signals containing the same information;

a transmit-side multiplexer for converting the first plurality of optical signals into a multiplexed fabric output signal for transmission to an input of an optically active switch;

wherein the receiving side comprises:

a receive-side demultiplexer for receiving a multiplexed fabric input signal from an output of an optically active switch and converting the multiplexed fabric input signal into a second plurality of optical signals;

a photodetector for converting the second plurality of optical signals into a second electronic signal; and

a receive-side output to send the second electronic signal to any one of:

a transmit side input of the second architecture part, or

An output device, via the client portion and the client port.

It should be noted that the term "port" in the above does not necessarily refer to a physical hole or socket. In contrast, a "client port" and a "fabric port" of a leaf switch refer to the area of the leaf switch that is responsible for preparing or processing signals to be sent to/received from the client and fabric, respectively. The leaf switch according to the second aspect of the invention is capable of acting as an intermediate leaf switch in that the second electronic signal may be sent to another fabric port (which may be referred to herein as a "fabric portion") to be retransmitted along a different dimension, or to a client port for transmission to an output device.

Thus, the second fabric port is preferably capable of performing the same function as the first fabric portion and has equivalent features. The second fabric port may be substantially identical to the first fabric port.

The leaf switch according to the second aspect of the invention provides the functionality required for building the scalable multi-dimensional optoelectronic switch of the first aspect of the invention. Due to the modulation of the first electrical signal, most of the data transmission can be done in the optical domain rather than in the electrical domain. This means that data can be transmitted at high data rates and over long distances with lower power and power consumption than in the case of the electronic domain. Furthermore, the use of the optical domain enables the use of Wavelength Division Multiplexing (WDM). Another important advantage is bit rate independence, where switch plane data operates at packet rate rather than bit rate.

The transmit side of each leaf switch preferably includes a transmit side packet processor configured to receive the first electronic signal in the form of a packet having a header containing destination information. In addition to the data itself, the information included in the packet may include information related to the destination of the packet (e.g., client port/destination leaf switch). The packet header may include additional items of information including source and destination addresses, packet length, protocol version, sequence number, payload type, hop count, quality of service indicator, and other information.

The transmit side packet processor may be configured to perform packet fragmentation, wherein data packets having the same destination leaf switch are arranged into frames having a predetermined size, and wherein the data packets may be divided into a plurality of packet fragments arranged in a plurality of corresponding frames, and wherein a frame may contain data from one or more data packets. Each packet fragment may have a packet fragment header that includes information identifying at least the packet to which the packet fragment originally belongs so that the packet may be reconstructed after subsequent processing and transmission.

For example, consider the following case: wherein the packet processor is configured such that the frame payload size is 1000B and three packets of 400B, 800B and 800B are input into the leaf switch. If each of these packets were to be sent in separate frames, one packet per frame, this would represent an efficiency of (400 + 800 + 800)/3000 = 67%. However, by using packet fragmentation, the first frame may include 600B of the 400B packet and the first 800B packet, and then the second frame may include the remaining 200B of the second 800B packet and the first 800B packet. This results in 100% efficiency. The frames built by this process represent the data packets themselves, and therefore more fragmentation can occur at the intermediate leaf switches when the packets undergo more than one hop in order to reach the destination leaf switch. It should be noted that even in the case where the signal passes through the spine switch and the two optically active switches in order to reach the intermediate leaf switches, this is still only a single hop. In other words, a hop may be defined as the transmission of a signal from one leaf switch to another.

To maximize efficiency, the frames may not be subsequently processed (e.g., forwarded for conversion into the first plurality of optical signals) until the fill fraction of the frames reaches a predetermined threshold, preferably above 80%, more preferably above 90% and most preferably 100%. Or the packet may be sent for subsequent processing after a predetermined amount of time has elapsed. In this way, if a packet for a given leaf switch stops arriving at the packet processor, frames that are still below the threshold fill fraction may still be sent for subsequent processing, rather than stalling on the packet processor. The predetermined amount of time may be between 50 ns and 1000 ns, but is preferably between 50 ns and 200 ns. Most preferably, the time interval is about 100 ns. Thus, the transmit side packet processor may include or be associated with a transmit side memory for temporarily storing incomplete frames during construction of the frames. The elapsed time may vary depending on the flow demand; in general, the higher the rate of flow, the shorter the elapsed time will be, and the lower the rate of flow may result in an increased time interval.

When the packet processor is configured to perform packet fragmentation, the receive side may further include a receive side packet processor configured to reconstruct the original packet from the packet fragments when the original packet is spread over more than one frame. This may be done with reference to the packet fragment header described above. When a packet undergoes several separate divisions of successive intermediate leaf switches on its way from the source to the destination, the final reassembly of the packet by the receiver-side packet processor may be delayed until all components of the original packet reach the destination leaf switch. Thus, the receive-side packet processor may include or be associated with receive-side memory to temporarily store the components.

The transmit side may comprise a plurality of modulators, preferably optical modulators. The optical modulator may be a phase or intensity modulator, such as an electro-absorption modulator (EAM), a Franz-Keldysh (Franz-Keldysh) modulator, a modulator based on the quantum confined Stark effect, a Mach-Zehnder (Mach-Zehnder) modulator, and the plurality of modulators preferably includes 8 modulators. Each modulator may be associated with only a single light source, or may be illuminated by fewer light sources, where the light sources are shared between the modulators. Each modulator may be configured to receive an electronic signal from an input or transmit side packet processor and unmodulated light from a light source. By combining the two, the modulator generates a modulated light signal that has the same wavelength as the unmodulated light from the light source and carries the information carried by the original (first) electronic signal. This modulated optical signal may then be passed to a transmit side multiplexer. The light source preferably takes the form of a laser to generate a substantially monochromatic beam of light limited to a narrow band of wavelengths. To minimize losses, the modulator is preferably configured to have light with a wavelength in the C-band or L-band of the electromagnetic spectrum (i.e., from 1530 nm to 1625 nm). More preferably, the light has a wavelength falling within the C-band or "erbium window", i.e. from 1530 to 1565 nm.

The laser may be a fixed wavelength laser or a tunable laser. In an array of modulators, the light sources associated with each modulator should have different wavelengths of non-overlapping bandwidth in order to minimize crosstalk in the multiplexer. When the light source is a laser, the modulator may take the form of an electro-absorption modulator (EAM) that modulates the intensity of the laser light using a varying voltage to carry the information contained in the electronic signal. Using EAM means that only the intensity of the laser light is changed, not the frequency, and thus any change in the wavelength of the modulated optical signal is prevented.

In embodiments where the leaf switch includes multiple modulators, the transmit-side packet processor may also be configured to perform packet slicing, where frames (as constructed by the packet splitting process described above) or data packets are sliced into a first plurality of electrical signals. Each of the first plurality of electrical signals may then be sent to a different one of a plurality of modulators, whereby the first plurality of electrical signals is converted into a first plurality of optical signals. The photodetector (which may be a photodiode) may convert the second plurality of optical signals into a second plurality of electrical signals. More preferably, the receiving side may comprise a plurality of light detectors, which may be identical to each other or substantially identical to each other. The receive-side packet processor may be configured to recombine the second plurality of electronic signals representing the packet slice into a second electronic signal. By dividing a packet or frame into multiple slices before sending to another leaf switch, data can be sent using many different wavelengths that are multiplexed into a single optical link by a multiplexer. In this way several items of information can be sent in parallel and result in increased bandwidth and more efficient data transfer.

In the case where the transmission-side packet processor is configured to perform both packet slicing and packet division, the packet division step (i.e., formation of a data frame) is performed first, followed by slicing of the frame. Accordingly, at the destination (or intermediate) leaf switch that receives the signal, the packet processor recombines the second plurality of electronic signals (i.e., the packet slice) into a single second electronic signal before reconstructing the original packet from the frame.

After the splitting, frames are constructed, each containing data intended for a single leaf switch, which refers to the next leaf switch in the overall transmission, not necessarily the final destination leaf switch. After this, the data is converted into a first plurality of optical signals of mutually different wavelengths, which are multiplexed, in particular wavelength division multiplexed, by a transmission-side multiplexer to form a multiplexed architecture output signal. Preferably, the leaf switch is configured to operate in burst mode, wherein the leaf switch is configured to send a series of consecutive bursts to the multiplexed fabric output signal, each burst comprising packets and/or packet fragments from a single data frame, and such that each burst comprises only packets and/or packet fragments to be sent to a single leaf switch in their next hop. Each successive burst may include data frames to be sent to a different single leaf switch in their next hop. The pairs of consecutive bursts may be separated by a predetermined time interval which may be between 50 ns and 1000 ns, but is preferably between 50 ns and 200 ns. Most preferably, the time interval is about 100 ns. Preferably, all fabric ports connected to active switches in a single sub-array are configured to operate synchronously, i.e. each fabric port simultaneously sends bursts to the input of the optical active switch. In this way, the optically active switch can route each signal to a given spine switch in one switching action.

The transmit side packet processor may also be configured to perform error correction on incoming data packets. This may be done by means such as error detection and retransmission or Forward Error Correction (FEC). In addition, the leaf switch may also include management ports configured to perform fabric management processes including initialization, programming of routing/forwarding tables, fault reporting, diagnostics, statistical reporting, and metering.

When a packet needs to perform more than one hop in order to reach its destination leaf switch, the route can be derived entirely from a comparison between the coordinates of the source leaf switch and the destination leaf switch. For example, in a process known as dimension-ordered routing, a first hop may match first coordinates of the source and destination leaf switches, a second hop may match second coordinates of the source and destination leaf switches, and so on until all coordinates match, i.e., until the packet has been transmitted to the destination leaf switch. For example, in a four-dimensional network, if the source leaf switch has coordinates (a, b, c, d) and the destination leaf switch has coordinates (w, x, y, z), then the dimension ordering route may be: (a, b, c, d) → (w, b, c, d) → (w, x, c, d) → (w, x, y, d) → (w, x, y, z). At any point along the route, the packet processor may compare the coordinates of the source leaf switch with the coordinates of the destination leaf switch and determine which coordinates have not yet matched. The route along the non-matching direction will then be decided, for example, using the lowest index or the highest index.

The optically active switch or optical circuit switch of the present invention may be a mach-zehnder interferometer (MZI) based arrangement and, more particularly, may take the form of an MZI cascaded switch. The MZI cascaded switch comprises a plurality of MZIs, each MZI having: two arms split at an input coupler, wherein the two arms feed separate paths into an output coupler where the two arms recombine; and two output sections. Multiple MZIs are preferably arranged to provide a path from each input to each output of the MZI cascaded switch. To the greatest possible extent, the arms have the same length. Alternatively, where it is desired to have a default output, the arms may be unbalanced. Each MZI may comprise an electro-optic region at one or both arms, with the refractive index depending on the voltage applied to the region via one or more electrodes. Therefore, the phase difference of light traveling through the photoelectric area can be controlled by applying a bias voltage via the electrodes. By adjusting the phase difference and thus the interference generated at the output coupler, light can be switched from one output to the other of the MZI.

Consider the following: wherein each optically active switch hasK _i An input terminal andK _i an output, which may be composed of, for example, a plurality of 1 x 2 and 2 x 1 MZIs, arranged to provide a path from each input to each output. When in useK _i At 5 or more, the MZI cascaded switch is in useK _i Is connected to an input terminalK _i Full mesh of individual outputs is beneficial because of the full mesh requirement ¹ / ₂ K _i (K _i +1) A root optical fiber connecting all the input ends to all the output endsOutput, whereas the MZI cascaded switch only needs 2K _i A root optical fiber. Possibly by establishingK _i “1 × K _i Demultiplexer Tree "andK _i “K _i x 1 multiplexer tree "to create a tree withK _i = 2nMZI cascaded switches of individual inputs and outputs, wherein each tree comprisesn1 x 2 (demultiplexer) or 2 x 1 (multiplexer) switch of a stage, of which the firstkAt stage has2kA switch. By establishing on each side (K _i +1) By having trees and omitting internal connections, additional ports may be supported on each cascaded switch so that an input is not connected to an output that is connected to the same switch as itself. MZI cascaded switches such as this are mostly wavelength agnostic and are therefore able to switch the entire multiplexed fabric output signal from input to output without first requiring any demultiplexing/multiplexing at the input and output.

The spine switch may be an electronic active switch or an electronic packet switch, such as an electronic crossbar switch. Preferably, the spine switch may be an electronically shared memory type switch. An electronically shared memory type switch is an electronic crossbar switch that also includes memory. The presence of memory within the switch is advantageous because it means that when a bottleneck occurs at the electronically shared memory type switch (as described above), the switch can perform not only switching but also buffering, i.e., storing queues of packets. This means that the electronics on the packet processor on the leaf switch can be simplified.

In a preferred embodiment of the invention, each of the spine switches and each of the leaf switches may contain the same components for ease of manufacturing. In other words, all leaf switches and spine switches may be identical or substantially identical to each other. In this way, an optoelectronic switch according to the invention can be built by assembling a set of identical or substantially identical elements, and the functions of these elements can then be controlled using, for example, software. Spine switches are fully "fabric-oriented," i.e., they are not connected to external devices, and accordingly, they do not have client ports or client portions. However, it has to be emphasized that the physical structure of the switching elements is the same as the leaf switches, only that the use or functionality of the various (identical) components may be distinguished. Thus, a spine switch may comprise:

a first fabric portion and a second fabric portion connected to a first fabric port and a second fabric port, respectively, to process signals and communicate with a leaf switch or an optically active switch, the first fabric portion having a transmit side and a receive side,

wherein the transmitting side comprises:

a transmit side input for receiving a first electronic signal carrying information, the information comprising information about a destination leaf switch of the first electronic signal, the first electronic signal received from an output of the second fabric portion;

a modulator for converting the first electronic signal into a first plurality of optical signals containing the same information;

a transmit-side multiplexer for converting the first plurality of optical signals into a multiplexed fabric output signal for transmission to an input of an optically active switch;

wherein the receiving side comprises:

a receive-side demultiplexer for receiving a multiplexed fabric input signal from an output of an optically active switch and converting the multiplexed fabric input signal into a second plurality of optical signals;

a photodetector for converting the second plurality of optical signals into a second electronic signal; and

a receive-side output to send the second electronic signal to a transmit-side input of the second architectural portion.

Where compatible, the optional features set forth above with reference to leaf switches apply equally to spine switches.

As discussed above, the spine switch may be some sort of electronic switch. Since the output from the optically active switch is an optical signal, these signals must be converted to electrical signals that are then switched using the spine switch. Thus, an optical-to-electronic (herein, "O/E") converter may be positioned at the inputs of the spine switches in the sub-arrays, and preferably at each input, for converting signals from, for example, a fabric ingress cluster of optically active switches. Similarly, in those embodiments having a second set of optically active switches, for example, in a fabric egress cluster, an electronic-to-optical (herein, "E/O") converter may be positioned at the output of the spine switch and preferably at each output. Between these O/E and E/O conversions, signals may be exchanged in the form of electronically exchanged signals. It is preferable to have O/E and E/O converters at the spine switch rather than at the optically active switch because, as previously discussed, it is advantageous to transmit the signals optically rather than electronically. The O/E converter may include a photodetector or a plurality of photodetectors, each of the plurality of photodetectors corresponding to an input of the spine switch.

In embodiments employing wavelength division multiplexing, the signals may first need to be demultiplexed before they can be electronically switched. Thus, the O/E converter may comprise a demultiplexer for demultiplexing the multiplexed fabric output signals (into, for example, a first plurality of optical switching signals) before converting the demultiplexed signals into electronic switching signals, whereby the signals are switched by the spine switch. The E/O converter may then convert the switched electronic switching signals into a corresponding second plurality of optical switching signals, which may then be multiplexed into a single signal to form a multiplexed fabric input signal for transmission to the optically active switch. Thus, the E/O converter may also include a multiplexer. In a preferred embodiment, the electronic active switch may be configured to temporarily store a queue of packets or data frames in the event that a request associated with the packet or frame cannot be satisfied.

Any or all of the multiplexer, transmit side multiplexer, demultiplexer and receive side demultiplexer preferably take the form of an Arrayed Waveguide Grating (AWG), the AWG being a passive device. AWGs allow multiple optical signals of different lengths to be carried along a single fiber. Since the wavelengths of the multiple modulated optical signals produced by the modulators are all different, the multiplexed architectural output signals produced by the AWG experience little crosstalk because the different wavelengths of light interfere only linearly. Alternatively, instead of an AWG, the multiplexed signal may be broadcast to a number of wavelength selective filters, each tuned to receive the wavelength of one of the desired split signals.

Bandwidth is a consideration in switching systems such as the switch of the present invention. In the following discussion, "bandwidth" is used to refer to the maximum data transmission rate that a particular portion is capable of achieving, and is typically measured in gigabits per second (abbreviated herein as "Gbps"). In particular, it is important to ensure that there is bandwidth conservation on a local and global scale. To ensure that more data is unlikely to enter a given leaf switch at a given time (i.e., causing a localized bottleneck at the leaf switch) than can be simultaneously transmitted away from the leaf switch, the total bandwidth of the client ports on the leaf switch does not exceed the total bandwidth of the fabric ports on the same leaf switch. More preferably, the total bandwidth of the fabric ports on a leaf switch exceeds the total bandwidth of the client ports on the same leaf switch, and most preferably, the bandwidth of each fabric port on a leaf switch exceeds or equals the total bandwidth of all client ports on the leaf switch. In this way, local bottlenecks caused by an unexpectedly large amount of input data from multiple client ports all pointing to the same fabric port on the same leaf switch may be avoided. In particular, this allows all signals to be multiplexed together for subsequent transmission in a non-blocking manner. In the above-described embodiments of the present invention, the leaf switches and spine switches are connected via optically active switches. However, it is also contemplated that other types of switches may be used instead of optically active switches connected between spine switches and leaf switches. For example, these switches may be referred to as intermediate switches. These intermediate switches may be electronic active switches, such as electronic packet switches, and may have the same components as the electronic active switches described above with reference to the spine switch. Of course, embodiments in which the intermediate switch is an optically active switch have been discussed in detail above. In some embodiments, each of the intermediate switches and each of the spine switches and/or leaf switches may have the same components. In embodiments that include these intermediate switches, the definition of leaf switches and spine switches only differs in that the "optically active switches" may be replaced by "intermediate switches". An intermediate switch may be defined as a switching element that includes:

a first fabric portion and a second fabric portion connected to a first fabric port and a second fabric port, respectively, to process signals and communicate with a spine switch or a leaf switch, the first fabric portion having a transmit side and a receive side,

wherein the transmitting side comprises:

a transmit side input for receiving a first electronic signal carrying information, the information comprising information about a destination leaf switch of the first electronic signal, the first electronic signal received from an output of the second fabric portion;

a modulator for converting the first electronic signal into a first plurality of optical signals containing the same information;

a transmit-side multiplexer for converting the first plurality of optical signals into a multiplexed fabric output signal for transmission to an input of a spine switch or a leaf switch;

wherein the receiving side comprises:

a receive-side demultiplexer for receiving a multiplexed fabric input signal from an output of a leaf switch or spine switch and converting the multiplexed fabric input signal into a second plurality of optical signals;

a photodetector for converting the second plurality of optical signals into a second electronic signal; and

a receive-side output to send the second electronic signal to a transmit-side input of the second architectural portion.

All of the optional features set forth above with respect to spine switches and leaf switches may also be applied to intermediate switches. Furthermore, all of the optional features set forth above with respect to an optoelectronic switch including an optically active switch may also be applied to an optoelectronic switch including an intermediate switch.

Further optional features of the invention are described below with reference to the drawings.

Brief description of the drawings

These and other features and advantages of the present invention will be appreciated and understood with reference to the specification, claims, and drawings, in which:

fig. 1 through 4 are schematic diagrams illustrating connections between leaf switches, spine switches, and optically active switches in a given sub-array according to three different embodiments of the present invention. The connections are shown in only a single dimension.

Fig. 5A/5B are schematic diagrams showing line cards and fabric cards, respectively.

Fig. 6 is a schematic diagram showing connections between leaf switches, spine switches, and optically active switches in a two-dimensional array of switching elements, depicting connections in two dimensions.

Fig. 7 is a schematic diagram showing an alternative representation of the connections between leaf switches, spine switches, and optically active switches in one dimension in a given sub-array. The figure highlights the difference in the "up" and "down" directions and shows the architecture and the arrangement of leaves and ridges on the line cards.

Figures 8, 9 and 10 show partially folded representations of a 2D optoelectronic switch.

FIG. 11 shows an example of a leaf switch that may be used in embodiments of the present invention.

Figure 12 shows the relationship between various quantities and parameters worth noting in an optoelectronic switch according to the present invention.

Detailed Description

The detailed description set forth below in connection with the appended drawings is intended as a description of exemplary embodiments of optoelectronic switches provided in accordance with the present invention and is not intended to represent the only forms in which the present invention may be constructed or utilized. This description sets forth features of the invention in connection with the illustrated embodiments. As indicated elsewhere herein, like element numbers are intended to designate like elements or features.

In the optoelectronic switch according to the invention, the leaf switches in each sub-array are connected by a spine switch and an optically active switch. An example of connections in a given sub-array is shown in fig. 1, where the cardinality of leaf switches (i.e., fabric ports and client ports) and spine switches (fabric ports only) is four, i.e., each has four inputs and four outputs. Here, it has to be emphasized that the eight leaf switches 1 to 8 on the left-hand side of the figure are identical to the eight leaf switches 1 to 8 on the right-hand side of the figure. They are each shown twice (i.e. in the "unfolded" configuration) because the connections are not bidirectional, and showing them in the "folded" configuration would be confusing. In this example, there are four spine switches shown along the center of the figure.

Most importantly for the present invention, there is also a group of eight fabric ingress optically active switches 1A to 8A, which in this embodiment are optical circuit switches and are referred to herein as "OCS", and a group of eight fabric egress OCS 1B to 8B. All OCS have a base of two. Unlike leaf switches, OCS 1A to 8A are different from OCS 1B to 8B. When connected as shown in fig. 1, the leaf switches, spine switches, and OCS form a 5-layer Clos network, with five layers:

1. leaf switches 1 to 8

2. OCS 1A to 8B

3. Spine switch

4. OCS 1B to 8B

5. Leaf switches 1 to 8.

The layout of the optoelectronic switch as shown in figure 1 will now be discussed in more detail in order to demonstrate the meaning of some of the terms used in the section "summary of the invention" above, with particular attention to the composition of the "cluster". In fig. 1, eight leaf switches 1 to 8 are arranged into four clusters, each cluster containing two leaf switches, i.e., {1, 2}, {3, 4}, {5, 6}, {7, 8 }. Each of these leaf switch clusters is associated with a cluster of fabric ingress OCS 1A through 8A, as shown in table 1 below.

Cluster of leaf switches	Clustering of fabric-entry OCS
		1、2	1A、2A
3、4	3A、4A
		5、6	5A、6A
7、8	7A、8A

Table 1.

Each leaf switch in the cluster is connected to each fabric ingress OCS in the associated cluster, e.g., leaf switch 1 is connected to fabric ingress OCS 1A and 2A, and leaf switch 2 is connected to fabric ingress OCS 1A and 2A. The same applies to the constituent leaf switches in each cluster. Still focusing on a single cluster, it can be seen in fig. 1 that fabric ingress OCS 1A is connected to spine switches 1 and 2, and fabric ingress OCS 2A is connected to spine switches 3 and 4. Thus, in a given cluster of fabric ingress OCS, no two fabric ingress OCS are connected to the same spine switch. In other words, in a given cluster of fabric ingress OCS, each of the fabric ingress OCS is connected to a different set of disjoint spine switches. In fig. 1, the same is true for all clusters of fabric ingress OCS.

There is a corresponding cluster arrangement of fabric egress OCS 1B to 8B. The associations between leaf switches 1 to 8 and fabric egress OCS 1B to 8B are shown in table 2 below.

Cluster of leaf switches	Clustering of fabric-egress OCS
		1、2	1B、2B
3、4	3B、4B
		5、6	5B、6B
7、8	7B、8B

Table 2.

As will be appreciated from the symmetrical arrangement shown in fig. 1, the connections between fabric egress OCS 1B to 8B and spine switches 1 to 4 are the same as those of fabric ingress OCS 1A to 8A. Fabric egress OCS 1B connects spine switches 1 and 2 and fabric egress OCS 2B connects spine switches 3 and 4. Thus, in a given cluster, no two fabric egress OCS are connected to the same spine switch. In other words, in a given cluster of fabric-egress OCS, each of the fabric-egress OCS is connected to a different disjoint subset of spine switches. In fig. 1, the same is true for all clusters that fabric-egress OCS.

Fig. 2, 3 and 4 show similar Clos network arrangements of switching elements within a given array, again including clusters of leaf switches with associated fabric-ingress and fabric-egress OCS clusters. The description of fig. 1 still applies to these drawings, and so these drawings will not be described in their entirety herein, with only a salient difference.

In fig. 2, the radix of the leaf switches and spine switches is six, and the radix of the fabric ingress and fabric egress OCS is two. Table 3 below lists associations between clusters of leaf switches, clusters of fabric-ingress OCS, and clusters of fabric-egress OCS.

Cluster of leaf switches	Clustering of fabric-entry OCS	Clustering of fabric-egress OCS
			1、2	1A、2A、3A	1B、2B、3B
3、4	4A、5A、6A	4B、5B、6B
			5、6	7A、8A、9A	7B、8B、9B
7、8	10A、11A、12A	10B、11B、12B
			9、10	13A、14A、15A	13B、14B、15B
11、12	16A、17A、18A	16B、17B、18B

Table 3.

Each of leaf switches 1 and 2 is connected to each of fabric egress OCS 1A to 3A. Then fabric ingress OCS 1A is connected to spine switches 1 and 2, fabric ingress OCS 2A is connected to spine switches 3 and 4, and fabric ingress OCS 3A is connected to spine switches 5 and 6. The same pattern applies to each of the clusters of leaf switches/fabric ingress OCS. The same applies to the cluster of fabric-egress OCS.

In fig. 3, the radix of the leaf switches and spine switches is four, as is the radix of the fabric ingress and fabric egress OCS. Table 4 below lists associations between clusters of leaf switches, clusters of fabric-ingress OCS, and clusters of fabric-egress OCS.

Cluster of leaf switches	Clustering of fabric-entry OCS	Clustering of fabric-egress OCS
			1、2、3、4	1A、2A	1B、2B
5、6、7、8	3A、4A	3B、4B
			9、10、11、12	5A、6A	5B、6B
13、14、15、16	7A、8A	7B、8B

Table 4.

Each of leaf switches 1 through 4 is connected to each of fabric egress OCS 1A and 2A. Then fabric ingress OCS 1A is connected to spine switches 1 through 4 and fabric ingress OCS 2A is connected to spine switches 5 through 8. The same pattern applies to each of the clusters of leaf switches/fabric ingress OCS. The same applies to the cluster of fabric-egress OCS.

Yet another example is shown in fig. 4, where the radix of the leaf switches and spine switches is four, the radix of the OCS is two, and the sub-array contains eight leaf switches. This figure also highlights how various components may be positioned on the line cards and fabric cards.

In fig. 1 to 4, a one-dimensional arrangement is shown. However, the present invention is by no means limited to one-dimensional arrangements such as this. For example, the arrangement of connections between the various switching elements shown in fig. 1-4 may be used to interconnect switching elements of a single sub-array of which a given leaf switch in a higher dimensional array is part. For example, consider a two-dimensional arrangement of leaf switches in rows and columns (in which case each row and each column represents a sub-array). Then the leaf switches in the sub-array corresponding to e.g. a given row may be connected as shown in fig. 1 to 4. In a preferred embodiment, for topological regularity, each of the sub-arrays in the array associated with a given dimension may be connected in the same manner, or all sub-arrays in the entire array may be connected.

In fig. 6, a two-dimensional array of leaf switches is shown, where the leaf switches and spine switches are organized with optically active switches on line cards and fabric cards, respectively. Fig. 5A and 5B show the arrangement of components on the line cards and fabric cards, respectively, in more detail. These will be described before discussing the higher levels of the overall array.

Fig. 5A shows a line card. The line cards include two leaf switches L1 and L2, and two (fabric ingress) optically active switches, in this case optical circuit switches O1 and O2. Each of OCS O1 and O2 is associated with a different dimension (denoted here as D1 and D2, where D1 is the horizontal dimension and D2 is the vertical dimension). Each of the leaf switches L1 and L2 has a connection to each of OCS O1 and O2 to facilitate switching in two dimensions. The exchange at D1 is the same as the exchange at D2, and therefore only the exchange at D1, i.e. only the right-hand bundle or the circled connected "D1 bundle" will be considered in this description. In particular, the D1 bundle contains two outputs from OCS O1 and two inputs from the fabric that are atIs not provided withPassing OCS O1 and O2 (one each) to L1 and L2.

FIG. 5B illustrates an architecture card. The fabric card includes two spine switches (or AOBMs) S1 and S2, and four (fabric egress) optically active switches. It should be noted that in the arrangement shown in fig. 6, and indeed all arrays have more than one dimension, components located on a given fabric card are associated with exchanges in only one dimension. This is illustrated in more detail with reference to the high level architecture of fig. 6. Each spine switch has a radix of four, i.e., it has four inputs and four outputs. First consider the input. The inputs do not come from OCS O1 'to O4' on the fabric card, as these are the fabric egress OCS through which the signal exits the fabric rather than entering it. Consider first S1. In contrast, the inputs to the spine switches S1 and S2 correspond to (i.e., are connected to) the outputs in the D1 bundle of each of the line cards contained in the sub-arrays for which the fabric card in question is responsible for switching.

Four outputs of each of the spine switches S1 and S2 are connected to OCS O1 'to O4'. More specifically, each of the spine switches S1 and S2 has one output to each of the OCS O1 'to O4'. Each of the fabric egress OCS O1 'through O4' (each having two inputs, one for each spine switch S1 and S2) has two outputs. These two outputs correspond to (i.e., are connected to) the two inputs in the D1 bundle discussed above with reference to fig. 5A.

The same connection arrangement is in place for the exchange on D2, but these are not shown in the drawings. By arranging the switching elements in this manner, a five-layer Clos network is formed as described above that is capable of switching signals in a non-blocking manner from one leaf switch in a given sub-array to another leaf switch in the same sub-array.

Fig. 6 shows a diagram of the entire array, unlike fig. 5A and 5B, which show only a single line card and fabric card, respectively. The "star-like" nature of the topology can be appreciated from fig. 6. Consider one of the sub-arrays in the array, e.g., the top row of line cards in the direction of D1 as shown. Each of the line cards LC 1-LC 4 in a sub-array is connected to a central fabric card FC1, through which FC1 signals are exchanged. The same applies to the sub-array in the direction D2. To understand the swapping process in detail, it is useful to profile the path of an exemplary signal in the direction D1 or D2 from, for example, LC2 to LC12 that are not located in the same sub-array. First, assuming dimension-ordered routing as described above, signals will be exchanged first in the direction of D1. Therefore, in the process proceeding via FC1, a signal is transmitted from LC2 to LC 4. More specifically, a signal may originate at L1 in LC2, which travels from L1 to O1 (because the signal is first switched in the direction of D1), and then takes one of the connections in the D1 bundle to the input of S2, e.g., FC1, and then switching occurs within spine switch S2, and the signal is output to O4' via one of the outputs of S2. The signal then travels from O4' to the input of L1, e.g., LC4 (bundled via D1 of LC 4). Here, the signals are internally switched to the input of O2 connected to LC4, i.e. to the output of L1 of the fabric ingress OCS associated with switching over dimension D2. The signal is then sent from the D2 bundle to the input of S1, e.g., FC8, FC8 is a fabric card containing spine switches responsible for switching in sub-arrays containing both LC4 and LC 12. Subsequently, the signal is switched within S1 of FC8 to an output connected to L2 of, for example, FC12, and then the signal goes from the output to L2 which travels to FC 12. At this point, the signal has reached the desired leaf switch, and then the signal is transmitted to the client output of L2 and from there to an external device. This process only extends into three or more dimensions, whereby each line card has one or more OCS and another connection bundle associated with switching in the third dimension.

Fig. 7 shows an alternative representation of the one-dimensional case, highlighting the difference between the "up" and "down" runs of the signal as it traverses a given sub-array. As shown in the figure, an "up" trip refers to a signal traveling from a leaf switch through the fabric ingress OCS and into a spine switch. Instead, the "down" trip is from the spine switch through the fabric egress OCS and into the destination leaf switch. The OCS shown in the "up" portion is the fabric ingress OCS that is located on a line card with the leaf switch. In this example, the cluster of leaf switches contains two leaf switches, and the associated clusters of fabric-entry OCS each contain only one fabric-entry OCS. In the "down" portion, the OCS is a fabric egress OCS and is disposed on a fabric card with two spine switches (which are the same spine switches shown in the "up" portion).

Figure 8 shows a one-dimensional example of connections in an optoelectronic switch according to an embodiment of the present invention as described in more detail above. In fig. 8, the leaf switches are depicted only once in a partially collapsed configuration. It should be noted that this is not a true folded Clos network, because the fabric-in and fabric-out OCS are different, so the signal cannot just travel "up" and "down" along the same route (see fig. 7). Similarly, FIG. 9 shows a partially folded representation of a 2-dimensional optoelectronic switch according to the present invention. Fig. 10 shows an alternative unfolded representation of a 2-dimensional optoelectronic switch comprising an array of sixty-four leaf switches at the edges of the figure and two sets of sixteen spine switches in the center (one set associated with each dimension). The dashed lines represent connections in one dimension and the solid lines represent connections in the other dimension.

It should be noted that embodiments of the present invention are not limited to arrangements in which the leaf switches and spine switches have a radix of four and the OCS has a radix of two.

Fig. 11 shows a more detailed view of the fabric side F1 of a typical leaf switch 1 used in an embodiment of the invention. First, the architecture of the leaf switch 1 will be described, followed by a description of the path of the signal through the leaf switch 1. The framework side F1 is divided into two parts: a transmission side Tx and a reception side Rx. The transmit side Tx includes an array of packet processors PP-Tx, EAM MOD1, MOD2,. and MODQ, each of which receives input from one of the light source arrays LS1, LS2,. and LSQ. Each of the EAM arrays is connected to a single multiplexer MUX which then outputs its WDM signal to an optically active switch which may be considered to be the "fabric" of interconnections between all the leaf switches 1 of an optoelectronic switch implementing an embodiment of the present invention. The reception side Rx has a similar structure. More specifically, the receive side Rx includes a packet processor PP-Rx that receives inputs from an array of photodetectors PD1, PD 2. The demultiplexer receives an input from an optically active switch (not shown in fig. 11). The controller CTRL is also included in the leaf switch 1, and is not limited to the transmission side Tx or the reception side Rx. The controller CTRL is bidirectionally connected to two packet processors PP-Tx, PP-Rx and to an arbiter, which is illustrated by an arrow labeled AR.

At a higher level, it should be noted that all data transmission taking place in the left-hand side of the figure takes place in the electrical domain and all data transmission taking place in the right-hand side of the figure takes place in the optical domain, i.e. all data transmission takes place between the multiplexer MUX and the demultiplexer DEMUX.

Now, the routing of packets through the various components of the leaf switch 1 will be described. Containing information to be transmitted from the source leaf switch to the destination leaf switch. Specifically, information relating to the intended destination leaf switch is included. In the following description of the trip taken by a packet, it is assumed that all data associated with the packet has the same intended destination leaf switch.

The following process is performed in the electrical domain. The packet may be incident on the transmission side Tx of the leaf switch 1, for example, from a client side part connected to the client side of the leaf switch 1. Alternatively, packets may be received from the receive side Rx of the leaf switch 1 (i.e., the same leaf switch) via the integrated switch such that the packets may be forwarded to another leaf switch (not shown) for transmission to a different dimension. A packet incident on the transmit side Tx enters a packet processor PP-Tx where it is sliced into a first plurality Q of electronic signals in the form of packet slices, each electronic signal having the same destination leaf switch. Each of the electrical signals is then communicated to one of Q EAM MOD1, MOD 2. At this time, each of the electrical signals contains information corresponding to the data in the packet slice and information related to the destination leaf switch of the packet.

Consider now the package slices incident on MOD 1. MOD1 has two inputs: (a) electrical package slice, and (b) light of a given wavelength from a light source LS 1. The optical channels are selected to minimize crosstalk and it is relatively easy to manufacture the waveguides with high yield. Light channel spacing between 0.4 nm and 2 nm is preferred. The laser will have a narrow linewidth practical for this application and preferably no less than 1 KHz. In other configurations, the frequency resolution and spacing will depend on the finesse of the device and thus the passive components. If there are, for example, 8 wavelengths, the device may be fairly "coarse", but if more wavelengths are used, higher specifications would be required.

MOD1 then modulates the light from light source LS1 to carry the information contained in the package slices in order to produce an optical signal with a given wavelength. From this point of view, the data transmission is in the optical domain. Each modulator operates similarly to produce a first plurality of Q optical signals. The Q slices of optical packets from each of EAM MOD1, MOD2, the. Each of the Q optical signals has a different wavelength and thus cross-talk between the signals is minimized. The multiplexed signal forming the multiplexed fabric output signal is then passed to the optically active switch. The optical signals generated in the leaf switches 1 are then transmitted via the fabric to their destination leaf switches or intermediate leaf switches and routed to the destination leaf switches.

For purposes of this description, reference will continue to be made to fig. 11, but in normal use, the source and destination leaf switches will not be the same leaf switch. However, the source and destination leaf switches may be the same leaf switch, e.g., for testing purposes. However, the source and destination leaf switches will be substantially identical to each other, and therefore the description based on fig. 11 applies as well. The optical multiplex architecture input signal from the optical active switch is incident on a demultiplexer DEMUX located on the receive side Rx of the leaf switch 1. The multiplexed architecture input signal is demultiplexed by a demultiplexer DEMUX into a second plurality Q of optical signals equivalent to those combined at the multiplexer MUX on the source leaf switch 1. The Q demultiplexed signals are then incident on each of the arrays of photodetectors PD1, PD 2. In the photodetection, the demultiplexed signal is converted back to a second plurality of Q electrical signals that also contain the information contained in the original packet slice. The electrical signals are then transmitted to the packet processors PP-Rx, where they are recombined into the original packets incident on the packet processors PP-Tx of the source leaf switch 1, using the information contained in the headers of the packet slices.

Figure 12 illustrates a mathematical relationship between various quantities involved in an optoelectronic switch, in accordance with an embodiment of the present invention, wherein:

•Lnumber of dimensions (= dimension)

•R ₁ Radix of leaves and spines

•R ₂ Cardinality of many leaves per cluster, number of spines per cluster

•C= client port per leaf, parallel spine

•F= architectural port per leaf

•U= unused ports per leaf

•N= total number of client ports

•P ₁ =Total number of leaves

•P ₂ =Total number of ridges

•P ₃= total number of OCS

•D= diameter

•B= halved bandwidth.

Although exemplary embodiments of optoelectronic switches have been described and illustrated herein with particularity, many modifications and changes will become apparent to those skilled in the art. It should therefore be understood that optoelectronic switches constructed in accordance with the principles of the present invention may be embodied other than as specifically described herein. The invention is also defined in the following claims and equivalents thereto.

Claims (32)

1. For transmitting optical signals from input means to output meansLA optoelectronic switch, the optoelectronic switch comprising:

multiple leaf switches, each leaf switch having a radixRAnd are prepared byLDimension array arrangement, wherein each dimensioniHave corresponding sizeR _i ，i = 1、2、...、LEach leaf exchangeThe machines having associations with co-ordinatesLTuple (x ₁、...、x _L ) Giving it a relation toLA location of each of the dimensions;

wherein each leaf switch isLMembers of a sub-array, theLEach of the sub-arrays and theLDifferent ones of the dimensions are associated and include:

a plurality ofR _i Leaf switches with co-ordinates only in the second placeiDimensionally differentiated, each leaf switch having means for connecting to an input or output deviceCIndividual client ports and for connection to a spine switchFAn architecture port;

a plurality ofS _i Spine switches each having the fabric ports for connecting to the leaf switchesRA plurality of fabric ports, and

wherein in a given sub-array, each leaf switch in the sub-array is connected to each spine switch via an optically active switch;

wherein each of the spine switches and each of the leaf switches contain the same components;

wherein each spine switch comprises:

a first fabric portion and a second fabric portion connected to a first fabric port and a second fabric port, respectively, to process signals and communicate with a leaf switch or an optically active switch, the first fabric portion having a transmit side and a receive side,

wherein the transmitting side comprises:

a transmit side input for receiving a first electronic signal carrying information, the information comprising information about a destination leaf switch of the first electronic signal, the first electronic signal received from an output of the second fabric portion;

a modulator for converting the first electronic signal into a first plurality of optical signals containing the same information;

a transmit-side multiplexer for converting the first plurality of optical signals into a multiplexed fabric output signal for transmission to an input of an optically active switch;

wherein the receiving side comprises:

a receive-side demultiplexer for receiving a multiplexed fabric input signal from an output of an optically active switch and converting the multiplexed fabric input signal into a second plurality of optical signals;

a photodetector for converting the second plurality of optical signals into a second electronic signal; and

a receive-side output to send the second electronic signal to a transmit-side input of the second architectural portion.

2. The optoelectronic switch of claim 1, wherein the number of leaf switches is greater than the number of spine switches in a given sub-array.

3. The optoelectronic switch of claim 1 or claim 2, wherein the leaf switches are divided into a plurality of clusters, each cluster containing a plurality of leaf switches.

4. The optoelectronic switch of claim 3, wherein:

each leaf switch cluster is associated with a cluster of one or more fabric ingress optical active switches to form a line card assembly,

each leaf switch in the cluster is connected to each fabric ingress optically active switch in the line card assembly, and

the fabric ingress optical active switch is arranged such that optical signals pass through the fabric ingress optical active switch during transmission from the leaf switch to the spine switch.

5. The optoelectronic switch of claim 4, wherein in a given sub-array:

each spine switch is connected to a fabric ingress optically active switch in a line card assembly positioned in the sub-array, and

no more than one fabric ingress optically active switch in the line card assembly is connected to a given spine switch.

6. The optoelectronic switch of claim 5, wherein within a line card assembly there is aLGroups of different fabric ingress optically active switches, each group configured to transmit optical signals within a respective sub-array containing the line card assemblies, each of the sub-arrays being in communication with the sub-arrayLRespective ones of the dimensions are associated.

7. The optoelectronic switch of claim 4, wherein each line card assembly is positioned on a respective line card.

8. The optoelectronic switch of claim 7, wherein the line card includes an arbiter for controlling a path of signals through a fabric ingress optical active switch included in the line card assembly positioned on the line card.

9. The optoelectronic switch of claim 8, wherein the line card comprises a plurality of arbiters, each arbiter configured to control a path of signals through a respective fabric ingress optical active switch.

10. The optoelectronic switch of any one of claims 4 to 6, wherein:

each leaf switch cluster is further associated with one or more clusters of fabric egress optically active switches to form a fabric egress arrangement,

each leaf switch in a given cluster is connected to each fabric egress optically active switch in the fabric egress arrangement, and

the fabric egress optical active switch is arranged such that optical signals pass through the fabric egress optical active switch during transmission from the spine switch to the leaf switch.

11. The optoelectronic switch of claim 10, wherein in a given sub-array:

each spine switch is connected to a fabric egress optically active switch positioned in a fabric egress arrangement in the sub-array, and

no more than one fabric egress optically active switch in the fabric egress arrangement is connected to a given spine switch.

12. The optoelectronic switch of claim 11, wherein the spine switch and the fabric egress optically active switch are positioned on a fabric card in a given sub-array.

13. The optoelectronic switch of claim 12, wherein the fabric card comprises an arbiter for controlling a path of signals through a fabric egress optical active switch positioned on the fabric card.

14. The optoelectronic switch of claim 13, wherein the fabric card comprises a plurality of arbiters, each arbiter configured to control a path of signals through a respective fabric egress optical active switch on the fabric card.

15. The optoelectronic switch of claim 10, wherein the number of fabric egress optical active switches is the same as the number of fabric ingress optical active switches in a given sub-array.

16. The optoelectronic switch of claim 15, wherein in a given sub-array, the leaf switches, spine switches, fabric ingress optically active switches, and the fabric egress optically active switches are arranged in a five-stage Clos network, wherein:

the first stage consists of the leaf switches;

the second stage consists of the fabric ingress optically active switch;

a third level is comprised of the spine switch;

the fourth stage consists of the fabric egress optical active switch;

the fifth stage consists of the leaf switches.

17. The optoelectronic switch of claim 1 or claim 2, wherein the number of inputs and the number of outputs on each optically active switch are the same.

18. The optoelectronic switch of claim 1 or claim 2, wherein the optically active switch is an optical circuit switch.

19. The optoelectronic switch of claim 18, wherein the optical circuit switch is a mach-zehnder interferometer (MZI) cascade switch comprising a plurality of MZIs, each MZI having: two arms split at an input coupler, wherein the two arms feed separate paths to an output coupler where the two arms recombine; and two output sections, the plurality of MZIs arranged to provide a path from each input to each output of the MZI cascaded switch.

20. The optoelectronic switch of claim 1 or claim 2, wherein the spine switch is an electronically active switch.

21. The optoelectronic switch of claim 20, wherein the electronic active switch is an electronic crossbar switch or an electronic shared memory type switch.

22. The optoelectronic switch of claim 1 or claim 2, wherein each leaf switch comprises:

a client port for connection to an input device or an output device and a client portion connected to the client port;

a first fabric portion and a second fabric portion connected to a first fabric port and a second fabric port, respectively, to process signals and communicate with a spine switch or an optically active switch, the first fabric portion having a transmit side and a receive side,

wherein the transmitting side comprises:

a transmit-side input to receive a first electronic signal carrying information, the information comprising information about a destination leaf switch of the first electronic signal, the first electronic signal received from any one of:

an output of the second architecture part, or

An input device via the client portion and the client port;

a modulator for converting the first electronic signal into a first plurality of optical signals containing the same information;

a transmit-side multiplexer for converting the first plurality of optical signals into a multiplexed fabric output signal for transmission to an input of an optically active switch;

wherein the receiving side comprises:

a receive-side demultiplexer for receiving a multiplexed fabric input signal from an output of an optically active switch and converting the multiplexed fabric input signal into a second plurality of optical signals;

a photodetector for converting the second plurality of optical signals into a second electronic signal; and

a receive-side output to send the second electronic signal to any one of:

a transmit side input of the second architecture part, or

An output device, via the client portion and the client port.

23. The optoelectronic switch of claim 22, wherein the transmit side of the first fabric portion comprises a transmit side packet processor configured to receive the first electronic signal in the form of a packet, the packet having a packet header containing the destination information.

24. The optoelectronic switch of claim 23, wherein the transmit-side packet processor is configured to perform packet fragmentation, wherein:

data packets having the same destination leaf switch are arranged into frames having a predetermined size;

a data packet may be divided into packet fragments over one or more frames; and is

A frame may contain data from one or more packets.

25. The optoelectronic switch of claim 24, wherein the receive side of the first fabric portion comprises a receive side packet processor configured to recreate an original data packet from the packet fragment when the packet is spread over more than one frame.

26. The optoelectronic switch of any one of claims 23 to 25, wherein:

the transmit side of the first architecture portion comprises a plurality of modulators;

the transmit-side packet processor is configured to perform packet slicing in which frames or packets are sliced into a first plurality of electronic signals; and is

After slicing, the transmit-side packet processor is configured to send each of the first plurality of electronic signals to a different one of the plurality of modulators, whereby the first plurality of electronic signals is converted into the first plurality of optical signals.

27. The optoelectronic switch of claim 26, wherein the receive side of the first fabric portion comprises a plurality of photo-detectors configured to convert the second plurality of optical signals into a second plurality of electronic signals, and further comprising a receive side packet processor configured to recombine the second plurality of electronic signals into the second electronic signal.

28. The optoelectronic switch of claim 23, wherein the transmit-side packet processor transmits frames and/or packet fragments in a series of consecutive bursts, each burst containing only packets and/or packet fragments having the same destination leaf switch, and wherein pairs of consecutive bursts are separated by a time interval.

29. For transmitting optical signals from input means to output meansLA optoelectronic switch, the optoelectronic switch comprising:

multiple leaf switches, each leaf switch having a radixRAnd are prepared byLDimension array arrangement, wherein each dimensioniHave corresponding sizeR _i ，i = 1、2、...、LEach leaf switch having an association with coordinatesLTuple (x ₁ 、...、x _L ) Giving it a relation toLA location of each of the dimensions;

wherein each leaf switch isLMembers of a sub-array, theLEach of the sub-arrays and theLDifferent ones of the dimensions are associated and include:

a plurality ofR _i Leaf switches with co-ordinates only in the second placeiDimensionally differentiated, each leaf switch having means for connecting to an input or output deviceCIndividual client ports and for connection to a spine switchFAn architecture port;

a plurality ofS _i Spine switches each having a connection for connectingConnected to said fabric port of said leaf switchRA plurality of fabric ports, and

wherein in a given sub-array, each leaf switch in the sub-array is connected to each spine switch via an intermediate switch;

wherein each of the spine switches and each of the leaf switches contain the same components;

wherein each spine switch comprises:

a first fabric portion and a second fabric portion connected to a first fabric port and a second fabric port, respectively, to process signals and communicate with a leaf switch or an optically active switch, the first fabric portion having a transmit side and a receive side,

wherein the transmitting side comprises:

a transmit side input for receiving a first electronic signal carrying information, the information comprising information about a destination leaf switch of the first electronic signal, the first electronic signal received from an output of the second fabric portion;

a modulator for converting the first electronic signal into a first plurality of optical signals containing the same information;

a transmit-side multiplexer for converting the first plurality of optical signals into a multiplexed fabric output signal for transmission to an input of an optically active switch;

wherein the receiving side comprises:

a receive-side demultiplexer for receiving a multiplexed fabric input signal from an output of an optically active switch and converting the multiplexed fabric input signal into a second plurality of optical signals;

a photodetector for converting the second plurality of optical signals into a second electronic signal; and

a receive-side output to send the second electronic signal to a transmit-side input of the second architectural portion.

30. The optoelectronic switch of claim 29, wherein each intermediate switch is an electronic active switch or an electronic packet switch.

31. The optoelectronic switch of claim 29, wherein each of the intermediate switches contains the same components as each of the spine switches and/or each of the leaf switches.

32. The optoelectronic switch of claim 31, wherein each intermediate switch comprises:

a first fabric portion and a second fabric portion connected to a first fabric port and a second fabric port, respectively, to process signals and communicate with a leaf switch or a spine switch, the first fabric portion having a transmit side and a receive side,

wherein the transmitting side comprises:

a transmit side input for receiving a first electronic signal carrying information, the information comprising information about a destination leaf switch of the first electronic signal, the first electronic signal received from an output of the second fabric portion;

a modulator for converting the first electronic signal into a first plurality of optical signals containing the same information;

a transmit-side multiplexer for converting the first plurality of optical signals into a multiplexed fabric output signal for transmission to an input of a spine switch or a leaf switch;

wherein the receiving side comprises:

a receive-side demultiplexer for receiving a multiplexed fabric input signal from an output of a leaf switch or spine switch and converting the multiplexed fabric input signal into a second plurality of optical signals;

a photodetector for converting the second plurality of optical signals into a second electronic signal; and

a receive-side output to send the second electronic signal to a transmit-side input of the second architectural portion.

Images