Optical network and node using time-slotted reception - Patents.com

Abstract

Proposed is a new data center (DC) network structure that addresses various limitations of Application Specific Integrated Circuit (ASIC) switches. In an optical network according to the present disclosure, a slight time domain restriction is introduced for transmissions from a network node to the same destination node. Data transmission and reception operate by means of a time slot system. A switching bandwidth of an ASIC switch has a bandwidth corresponding to average incoming traffic of a plurality of nodes, and a storage medium for storing and processing traffic exceeding this bandwidth is provided. An average bandwidth for a reception ASIC switch can also be set to a bandwidth obtained by reducing a maximum incoming bandwidth to a node by a factor F. Restricting the bandwidth of the reception ASIC switch also makes it possible to reduce the power consumption of the node, which is beneficial for the scalability of the network.

Description

[Background technology]

[0001] The present invention relates to an optical network, an optical node used therein, and an optical transmission method.

[0002] A data center is a general term for a facility that specializes in installing and operating computers (mainframes, minicomputers, servers, etc.) and data communication devices. Data center (DC) networks have grown rapidly to provide a wide range of services while keeping up with the massive increase in traffic volume. Despite this massive increase in traffic volume, the basic structure of the network has not changed significantly.

[0003] FIG. 1 is a diagram showing the configuration of a DC network of the prior art. A typical DC network is composed of multiple layers, and for example, as shown in FIG. 1, it has a three-layer DC network configuration. The three-layer DC network 100 includes, from the upper layer side, a switch 101, an optical link 102, a switch 103, an optical link 104, a ToR (Top Of Rack) switch 105, and a server 106. The switches 101 and 103 are respectively composed of upper and lower electrical switches and optical links 107 and 108 connecting them. Therefore, the DC network of FIG. 1 has eight hops between two ToRs. Optical-Electrical-Optical (OEO) conversion is performed over a wide area of ​​each electrical switch.

[0004] In such a DC network 100, data is transmitted as optical signals on links connecting network nodes each consisting of one or more servers. At each node, the optical signals are converted to electrical signals and switched by electrical switches, i.e., ASIC (Application Specific Integrated Circuit) switches. The switching capacity of ASIC switches continues to increase significantly, and currently, ASIC switches with a capacity of 12.8 Tb / s have been reported. [Non-Patent Document 1] White Rabbit project webpage, Internet<URL: https: / / www.ohwr.org / project / white-rabbit / wikis / home > [Non-Patent Document 2] K. Clark et al., “Sub-Nanosecond Clock and Data Recovery in an Optically-Switched Data Center Network,” 2018, post-deadline paper in ECOC 2018, Italy Summary of the Invention [Problem to be solved by the invention]

[0005] However, the expansion of the switching capacity of ASIC switches is expected to eventually reach a limit because it is becoming increasingly difficult to further reduce the size of CMOS (Complementary Metal Oxide Semiconductor) transistors, the building blocks of ASIC switches. Post-planar CMOS manufacturing processes are being sought to make CMOS transistors smaller, but such processes are technically challenging and very costly.

[0006] A more pressing issue is the extremely high data rate required for optical signals entering and leaving the ASIC switch. To accommodate the high data rates, co-packaging transceivers and ASIC switches has been attracting attention in recent years as an alternative to traditional pluggable transceivers. However, even with this advanced approach, the high power consumption and heat generation remain issues.

[0007] The heat generated by SAIC switches has already exceeded the limits of traditional air cooling, and water cooling is required to handle the heat dissipation of ASIC switches. Although water cooling can provide some power consumption margin for ASIC switches, it is not a long-term solution. Thus, faced with the significant increase in the data speed of optical signals in ASIC switches, the problems of spatial density and power limitations have become serious obstacles.

[0008] The fundamental solution to all the problems mentioned above is to avoid or reduce the frequency of optical-electrical-optical (OEO) conversions, which are inevitable for ASIC switches, and thus requires the introduction of optical switching into DC network structures.

[0009] The present invention has been made in view of the above problems, and its purpose is to propose a new DC network architecture that addresses the various limitations of ASIC switches. [Means for solving the problem]

[0010] To achieve this objective, one embodiment provides an optical network comprising an optical core section having a full mesh network configuration, and a plurality of nodes connected to the optical core section, the plurality of nodes being divided into a plurality of groups, with each group including up to m nodes, each of the plurality of nodes comprising an ASIC switch that switches electrical signals corresponding to optical signals received from the maximum m nodes and routes them to a plurality of servers, the ASIC switch having a switching capacity corresponding to the average incoming traffic of the plurality of nodes, and characterized in that during a receiving cycle period including a plurality of time slots, the ASIC switch is addressed by any node in the group to which the source node belongs only in a time slot associated with the group to which the source node belongs, and in one or more idle time slots subsequent to the plurality of time slots, no optical signals are received from any of the source nodes, and traffic exceeding the average incoming traffic is processed.

[0011] Another embodiment of an optical network is an optical network comprising an optical core section having a full mesh network configuration and a plurality of nodes connected to the optical core section, the plurality of nodes being divided into a plurality of groups, with each group containing up to m nodes, wherein each of the plurality of nodes is addressed by any node in the group to which the source node belongs only in a time slot associated with the group to which the source node belongs during a receiving cycle period including a plurality of time slots, and is an ASIC switch that switches electrical signals corresponding to optical signals received from the maximum m nodes and routes them to a plurality of servers, characterized in that the main switch has a switching capacity corresponding to the average incoming traffic of the plurality of nodes and operates in synchronization with the time slots, and an auxiliary switch has a switching capacity capable of handling traffic exceeding the average incoming traffic and operates regardless of the time slots.

[0012] The optical network described above also has an aspect of an invention regarding a network node within the optical network. Effect of the Invention

[0013] The optical network disclosed here simplifies the node configuration and reduces the capacity and power consumption of ASIC switches, while also supporting larger-scale optical networks with lower power consumption. [Brief description of the drawings]

[0014] [Figure 1] FIG. 1 illustrates a configuration of a DC network according to a conventional technique. [Diagram 2] FIG. 1 is a diagram showing a basic configuration of a DC network according to the present disclosure. [Diagram 3] FIG. 1 illustrates a problem with nodes at the periphery of optical networks in the prior art. [Figure 4] 1 is a diagram conceptually illustrating the configuration of an optical network and a receiving node according to the present disclosure. [Diagram 5] FIG. 13 is a diagram showing the receiver reduction effect obtained in the optical network of the present disclosure. [Figure 6] FIG. 2 is a diagram illustrating a data receiving operation at a receiving node in an optical network. [Figure 7] FIG. 13 is a diagram showing another example of the receiving operation of a network having a different group configuration. [Figure 8] 1 illustrates an example network and node specification according to the present disclosure. [Figure 9] 1A-1C illustrate different concepts of required reception bandwidth at a receiving node. [Figure 10] FIG. 13 is a diagram illustrating possible combinations of multiple nodes with different average numbers of connections. [Figure 11] FIG. 1 is a diagram illustrating the effective BW degradation between node pairs due to network partitioning. [Figure 12] FIG. 1 illustrates different bandwidths of traffic in a DC network. [Figure 13] FIG. 1 illustrates the concept of reduced bandwidth of a receive ASIC switch. [Figure 14] FIG. 13 is a diagram illustrating the relationship between the average bandwidth BWswitch_avg and the reception cycle period T. [Figure 15] FIG. 13 is a diagram illustrating an extended reception cycle period due to the introduction of a coefficient F. [Figure 16] FIG. 13 is a diagram showing the configuration of a node including a modified switch with the introduction of a factor F. [Figure 17] FIG. 1 is a diagram of a node that reduces delay in a modified switch. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] The following disclosure includes a DC network structure that realizes end-to-end optical transmission between desired node pairs with optical switching in the core and electrical switching only at the network periphery. The inventors have proposed a DC network structure that utilizes electrical switches more efficiently, which face performance limitations. In this specification, a new mechanism is proposed that flexibly addresses the capacity (bandwidth: BW) limitations in electrical switches with a new approach. First, a basic DC network and nodes that limit the bandwidth at the destination node by time slot operation are outlined. Then, a new mechanism is described that addresses the performance limitations of ASIC switches in a DC network with a basic configuration. [Optical network core configuration]

[0016] FIG. 2 is a diagram showing a basic configuration of a DC network according to the present disclosure. Although details will be described later, the DC network 1 includes a flat optical network 2, which is a core part of the entire network, and a SW unit 3, ToR 4, and a server 5, which are located in the periphery of the entire network. These peripheral elements constitute parts of nodes. In FIG. 2, for simplicity, one node including four ToRs is shown as one node, and two nodes are shown, but it should be understood that many other nodes are located in the periphery of the flat optical network 2. As already mentioned, optical switching is used in the flat optical network 2, which is the core part, and electrical switching is used only in the SW unit 3, which is the peripheral part, and on its peripheral side. The optical network structure of FIG. 2 is scalable and supports highly dynamic connections between any pair of nodes. Depending on the number of nodes included, the core part of the network can be implemented as a physical full mesh network or a full mesh-like network. The full-mesh network or full-mesh-like network assumed in the DC network of the present disclosure shown in Fig. 2 differs from the DC network 100 of the conventional technology shown in Fig. 1 in that only optical switching is used in the core section. The flat optical network 2 in Fig. 2 uses only optical switching without performing OEO conversion in the SW101 and 103 of each layer in Fig. 1.

[0017] The following disclosure reveals a novel and practical data receiving mechanism and associated hardware at each node, which replaces the numerous optical receiver units and parts of the complex high-capacity switching fabric previously required. The fundamental approach of the disclosed optical network architecture is to introduce a small time domain restriction on transmissions from a network node to the same destination node. In particular, the disclosed network operates in a time slot fashion. [Reception problems at the periphery of optical networks] To realize a physical full-mesh connection in the prior art N-node DC network shown in Figure 1, the number of transmitters, receivers, and bidirectional optical links is N x (N-1), which requires huge resources. In such a full-mesh or full-mesh-like network, a large-scale switching configuration is required at each node. In a full-mesh or full-mesh-like network with N nodes, any destination node in the network can be simultaneously addressed by any number of source nodes, up to N-1. Here, "addressing" is not limited to a source node identifying and designating a destination node as a communication destination in order to set up a communication link between nodes, but also includes actually setting up a communication link and carrying out communication.

[0018] Fig. 3 is a diagram illustrating the problem at nodes on the periphery of a conventional optical network. Fig. 3 conceptually illustrates a full-mesh optical network 140, which represents the flat optical network 1 shown in Fig. 2 as a collection of nodes. In the optical network 140, many nodes are interconnected. In Fig. 3, one node is depicted as having four neighboring nodes interconnected, which symbolically indicates that it is connected to all nodes in the optical network.

[0019] Here, attention is focused on one receiving node 141. The receiving node 141 has an interface unit 144 with the optical network side, and is further connected to the ToR 145. The interface unit 144 is equipped with a receiver 143 that receives multiple optical signals 142 arriving simultaneously. If the optical network 140 is composed of N nodes, the receiving node 141 needs a large number of ports and corresponding (N-1) receivers 143 to process all data communications arriving simultaneously. Furthermore, the interface unit 144 is equipped with a large-scale receiving ASIC switch (not shown). As mentioned at the beginning, a very high data rate is required for the optical signals inputted and outputted to the ASIC switch. It can be seen from FIG. 3 that a large number of receivers and a large-scale ASIC switch are required in all receiving nodes. In this way, nodes on the periphery of the flat optical network have the problem of the need for a huge number of receivers and the capacity limit of the ASIC switch. To solve these problems, the optical network disclosed in this disclosure proposes the introduction of a slight time domain restriction on data transmission from a network node to the same destination node. [Proposed network method]

[0020] A. Reducing the number of receivers FIG. 4 is a diagram conceptually illustrating the basic configuration of an optical network and a receiving node of the present disclosure. As in FIG. 3, FIG. 4 conceptually illustrates a full-mesh or full-mesh-like optical network 10. For simplicity, the optical network 10 is composed of N (e.g., 36) nodes, and as an example, data transmission between a source node, i.e., a source node 11, and a destination node, i.e., a receiving node 13 is considered. Naturally, in the following description, a source node can also be a receiving node, and the names are merely distinguished according to the operations and functions to be described. Therefore, it should be noted that a source node also has the functions and configurations described for a receiving node.

[0021] In the optical network 10, all N network nodes that are source nodes are divided into multiple groups (d), and a group to which each source node belongs is defined. In the example of Fig. 4, six groups, a first group (G1) to a sixth group (G6), are defined, and the source node 11 belongs to the first group 12.

[0022] Reception of data communication at any node, i.e. any receiving node, occurs separately in time slots of fixed duration T for each group of source nodes, as will be described later. That is, a receiving node can be addressed by any source node belonging to a group of source nodes only during the time slots assigned to that group. Specifically, receiving node 13 receives data from source node 11 during the time slots assigned to G1 to which source node 11 belongs. It is also addressed during the same time slots by the other five source nodes belonging to G1.

[0023] In Fig. 4, the nodes belonging to one group are shown as being close to each other for the sake of explanation, but in general, they are spatially distributed throughout the network. The receiving node receives data from different groups in succession for each time slot, and the cycle time for completing reception from all groups is d (number of divisions) × T (TS time). In the case of network division where the number of nodes in each group is uniform, the number of nodes m belonging to the same group is m = N / d.

[0024] As mentioned above, if a time limit is set for data reception at a receiving node, each receiving node must have m receivers to handle the maximum possible simultaneous incoming traffic. Here, the maximum transmission bandwidth (BW) of each node is defined as B outThe term "bandwidth" used in the following description means a transmission bandwidth that can be received or transmitted by a node, and may be understood as a transmission speed (transmission rate). It should be noted that the term "bandwidth" is a broad concept that means the capacity of communication resources for data transmission that is determined according to the modulation method and signal configuration of optical communication.

[0025] Usually, the bandwidth that can be sent from a node is the same as the bandwidth that can be received by a node. out is also the maximum receiving bandwidth. During the assigned time slot T, this B out However, a source node must wait one reception cycle before it can address the same receiving node again. For a sufficiently long observation time, the effective bandwidth between two nodes, or B out / d. In general, increasing the number of node groups d in the network reduces the effective BW between any pair of nodes. A new mechanism for increasing the effective BW is described below that addresses the performance limitations of the optical network of the present disclosure of FIG. 4 and the electrical switches at the receiving nodes.

[0026] By introducing slight time-domain constraints on data transmission from the above-mentioned source node to the same receiving node, the number of receivers can be significantly reduced in the nodes of the disclosed optical network of FIG. 4 compared to the nodes of the prior art configuration shown in FIG.

[0027] Referring again to FIG. 4, the details of the receiving node 13 are shown as a detailed diagram at the bottom, and will be described in comparison with the configuration of the receiving node according to the conventional technology shown in FIG. 3. In the node 13 in the optical network of the present disclosure, an arrayed waveguide grating (AWG) 14 that multiplexes optical signals from different wavelengths (d different groups) is added to the node front section that receives the input optical signal of the received data. Each of the AWGs 14 has at least d input ports, which is the number of groups, so that data from N nodes can be received simultaneously, and the front section of the node 13 is provided with m AWGs, which is the number of nodes in one group. The multiplexed received data that passes through the m AWGs 14 is supplied to an interface section 15 similar to the conventional technology and input to a receiver 16. Here, the number of receivers 16 in the interface section 15 is significantly reduced to 1 / m compared to the configuration of the conventional technology shown in FIG. 3.

[0028] Fig. 5 is a diagram showing the effect of reducing the number of receivers obtained by dividing the nodes into groups in the optical network of the present disclosure. In Fig. 5, the relationship between the number of receivers (Rx units) in a node and the total number of nodes is shown with the division number, i.e., the number of groups, as a parameter, and the number of receivers on the vertical axis is shown in logarithmic form in (a) and in linear form in (b). Regardless of the number of nodes, the number of required receivers can be significantly reduced as the division number increases.

[0029] Therefore, the optical network of the present disclosure can be implemented as comprising an optical core unit 2 having a full mesh network configuration, and a plurality of nodes connected to the optical core unit, the plurality of nodes being divided into a plurality of groups, with each group including a maximum of m nodes, and each of the plurality of nodes 13 being addressed by any node in the group to which the source node 11 belongs only in a time slot associated with the group to which the source node belongs. Also, each of the plurality of nodes can be implemented as comprising m arrayed waveguide gratings (AWGs) 14, the plurality of input ports of which receive corresponding optical signals from the maximum m nodes belonging to the same group, and the operating wavelengths of one or more source nodes among the maximum m nodes are set to match the operating wavelengths of the plurality of input ports, m receivers 16 connected to the output multiplexing ports of the AWG, and an ASIC switch 17 which switches and routes electrical signals from the m receivers to a plurality of servers 18.

[0030] Although the above description is of an invention of an optical network, it should be noted that the invention also has an aspect of an invention of a network node. For example, the invention can be embodied as a node connected to an optical core having a full mesh network configuration, the node being divided into a plurality of groups together with a plurality of other nodes connected to the core, with a maximum of m nodes being included in one group, and each of the maximum m nodes being addressed by any node in the group to which the source node belongs only in a time slot associated with the group to which the source node belongs.

[0031] Returning to the basic configuration of the receiving node in Figure 4, the interface unit 15 of the receiving node 13 includes a receiving ASIC switch 17, as in the configuration of the prior art. However, as will be described later, its switching capacity is significantly reduced compared to the prior art. The configuration of the interface unit in the receiving node is significantly simplified compared to the prior art in terms of the number of receivers and the capacity of the receiving ASIC switch. The operation of the receiving node and the reduction in the bandwidth of the receiving ASIC switch will now be described in more detail. B. Simplified receiving switching Let the maximum outgoing bandwidth of one node be B. out and the number of nodes in one group, m, the maximum BW that a node in the basic configuration of Figure 4 can receive in any time slot is m × B out This maximum BW is reduced by approximately 1 / d compared to the BW when data is received by completely asynchronous operation (e.g., optical packet switching) without using time slots. As shown in the configuration of the receiving node in Figure 4, the input data to the receiving node needs to be switched to the desired ToR switch 18. The receiving ASIC switch 17 is used for this switching.

[0032] 6 is a diagram for explaining the data receiving operation of a receiving node of a basic configuration in an optical network of the present disclosure. FIG. 6(a) shows a time slot configuration, and FIG. 6(b) is a diagram for explaining the switching operation when data is received in TS1. The ASIC switch 17 of the interface unit 15 in FIG. 6(b) has m ports on the input side facing the network, and a number of ports corresponding to the number of ToR switches 18 on the other output side. When multiple links are used for each ToR switch, there are a number of ports according to the multiple links.

[0033] The time slot configuration in FIG. 6(a) shows an example in which all the nodes in the optical network 10 shown in FIG. 4 are divided into six groups, and the m nodes belonging to each group receive data only in the corresponding time slot. For example, data from six source nodes belonging to the first group is received only in TS1 at the corresponding ports of each AWG as shown in FIG. 6(b). Data from m different source nodes is received in m AWGs 14-1 to 14-6, the number of which is the same as the number of nodes in one group. Therefore, in a specific time slot, each of the m input ports of the ASIC switch 17 can take in data received from different source nodes in the same group. That is, data can be received simultaneously from all m source nodes in the same group.

[0034] As shown in Fig. 6(b), to simplify the switching process of different groups, an AWG is connected to each port on the input side of the ASIC switch 17, and nodes of the same group are connected to different AWGs. During the assigned time slot, only the nodes of the authorized group are active, while all the nodes of the other groups do not transmit to the destination node.

[0035] For illustration purposes, assume that the same wavelength λ1 is used by all nodes of an active group (e.g., G1) in TS1. The input signal of wavelength λ1 is passively routed by multiple (m) AWGs towards ASIC switch 17, i.e., by connecting different nodes in the same group with ports of corresponding wavelengths of the m AWGs. In the next TS2, assume that all nodes in the next active group (G2) are transmitting on λ2. The input signal of λ2 from each node belonging to G2 is again passively routed by multiple (m) AWGs towards ASIC switch 17.

[0036] In general, nodes in the same group do not need to transmit at the same wavelength. The wavelengths used by nodes in different groups connected to the same AWG can be tuned to different wavelengths so that they do not conflict with each other. As explained above, if all nodes in the same group use the same wavelength, the AWG configuration at the receiving node can be standardized.

[0037] FIG. 7 is a diagram showing another example of the operation of reception switching in a network with a different group configuration. FIG. 7 shows (a) an example of reception in TS1 from G1, (b) an example of reception in TS2 from G2, and (c) an example of reception in TS16 from G16, in the case where the total number of nodes N is 512, the number of groups (number of divisions) d is 16, and the number of nodes in the group m is 32. In each diagram, the node numbers (1 to 512) are shown at the input of the AWG. In this example, the 32 nodes in the same group are shown to use the same wavelength, but the wavelength may be different for each AWG. By using 32 AWGs 14-1 to 14-32 in the receiving node, data can be received simultaneously from 32 different nodes belonging to the same group in the same time slot. In this configuration, it is sufficient to provide 32 receivers corresponding to the 32 AWGs in the interface section of the receiving node.

[0038] FIG. 8 shows an example network and node specification according to the present disclosure. It shows example parameters for a network divided into 16 and corresponding configuration with 16 time slots. Specifically, one time slot is 40 ns long, including 30 ns corresponding to a 400 Gb / s packet and 10 ns margin. One cycle is divided into 16 time slots, and 512 network nodes are divided into 16 groups. The number of nodes in one group is 32. The destination (receiving) node of each time slot is addressed only by the authorized group. The same source node can address the same destination node every 16 time slots. C. Reducing the BW of the receiving ASIC switch As mentioned in the explanation of FIG. 4, in the basic node of the optical network of the present disclosure, the capacity of the receiving ASIC switch is significantly reduced compared to the conventional technology. Specifically, the capacity of the receiving ASIC switch is set to a value that matches the average incoming traffic or slightly exceeds the average incoming traffic. In order to process the traffic that exceeds this switching capacity with some delay, a memory associated with the receiving ASIC switch is provided. Hereinafter, with reference to FIG. 9 and FIG. 10, it will be explained that even if the capacity of the receiving ASIC switch is reduced compared to the conventional technology node, the basic node of the flat optical network of the present disclosure can handle substantially sufficient traffic.

[0039] In order to reduce the switching load of the ASIC switch in the node, the inventors have classified the traffic input to the receiving node, in other words, the input bandwidth BW of the receiving node, into two categories with different requirements. The two traffic categories are (a) traffic that needs to be switched in real time without additional delay, and (b) traffic that does not need to be switched in real time by the receiving ASIC switch because it has already exceeded the processing capacity of the connected server. The average receiving BW is assigned to the traffic of (a), and the maximum receiving BW is assigned to the traffic of (b). In the following, we consider how the switching capacity of the receiving ASIC switch, i.e., the transmission bandwidth of the receiving ASIC switch, can be allocated to different BWs corresponding to the two categories of traffic.

[0040] In the case of maximum receiving BW that does not require real-time processing, introducing a storage medium (memory) for partially storing the received data of the receiving node ensures that the received data is not lost or needs to be retransmitted. The main advantage of introducing a storage medium is that it allows the switching capacity of the receiving ASIC switch to be reduced, at the expense of introducing an extra additional delay.

[0041] FIG. 9 is a diagram for explaining different concepts of the reception bandwidth required at a receiving node. In FIG. 9(a), the number of groups (the number of network divisions) d is set as a parameter (d=8, 16, 24, 32), and the bandwidth B that one node can receive is calculated. out The figure shows the relationship between the receiving BW normalized by and the number of nodes in the network, N. Here, the bandwidth of the receiving node, B out is the same as the maximum transmission bandwidth of the source node described in Figure 4. Taking a network with 512 nodes divided into 16 groups as an example, each receiving node should have 32 (=512 / 16) receiver units to accommodate the normalized maximum BW, which is 32 times the BW that the node itself can generate. If we assume that the normalized average receiving BW is 8, the switching capacity of the receiving ASIC switch is reduced by 8 / 32 compared to the case of maximum switching load. Figure 9 (b) is a diagram explaining the two receiving bandwidths mentioned above, and shows that 32 receivers are required at the front of the interface section to which the AWG output is connected in order to reliably receive all input data from 32 nodes simultaneously. On the other hand, it shows that the receiving ASIC switch in the interface section has memory to have an average receiving BW for 8 nodes and can perform real-time switching processing for 8 nodes.

[0042] Although the average reception BW may vary for each node, except for a limited number of specific nodes, the average reception BW can usually be set significantly lower than the maximum reception BW. This is due to the natural balance of traffic in the flat optical network 2 assumed in the DC network of the present disclosure shown in Fig. 2. This balance mechanism can be explained as the nature of traffic in the flat optical network as follows.

[0043] Figure 10 is a diagram explaining possible combinations of multiple nodes with different average number of connections. Here, we explain the nature of traffic assumed in a flat optical network. In a flat optical network with uniform connectivity and 100% traffic load, the average number of incoming connections per node is 1. From this state, when a node starts to receive more connections on average, the average number of incoming connections starts to deviate from the state of 1. At this time, in a flat optical network, this means that traffic that was being received by a certain node is directed to a new busy node. As more nodes become busy, the average number of connections of the remaining nodes naturally decreases.

[0044] In the example of Figure 10, we simulate an example network with 512 nodes and 100% traffic load to numerically elucidate the idea of ​​this balancing mechanism. To simplify the calculations without loss of generality, we classify the network nodes into three categories: (a) busy nodes with an average number of connections received Q (>1), (b) non-busy nodes with an average number of connections received 1, and (c) nodes that are not receiving data. The horizontal axis of Figure 10 shows the percentage of non-busy nodes with traffic of category (b) in %. The vertical axis shows the maximum percentage of busy nodes with an average number of connections Q greater than 1, corresponding to the traffic of category (a) mentioned above. Thus, the graph in Figure 10 shows the relationship between the percentage of nodes in category (b) (horizontal axis) that meet the condition of 100% traffic load and the percentage of nodes in category (a) (vertical axis) with the value of the average number of connections Q of busy nodes in category (a) as a parameter.

[0045] As shown in Fig. 10, when the value of the average number of connections Q of busy nodes increases to 2, 4, 6, and 8, the maximum percentage of busy nodes in category (a) with such average number of connections Q decreases significantly. Specifically, point A represents an extreme condition where the percentage of non-busy nodes in category (b) with one connection is 10%, and the percentage of nodes in category (c) with zero connections that are not receiving data is 79%. Even under such conditions, it can be seen that up to 11% of nodes (100-10-79=11) can have an average number of connections Q of 8. At this point A, a 100% traffic-loaded network is established with a node distribution state in which the majority (79%) of nodes are not receiving data, and the remaining nodes are a small number (10%) of non-busy nodes (Q=1) and a small number (11%) of nodes with a traffic-intensive average number of connections Q=8. In this example, it is strongly indicated that a receive ASIC switch with only a small capacity, such as the average incoming traffic, can be used sufficiently from the viewpoint of the maximum receive switching capacity of the receive ASIC switch. The traffic load simulation in Figure 10 shows that in a network with 512 nodes divided into 16 groups, the traffic can be adequately handled by setting the capacity of the receive ASIC switch to a normalized average receive BW=8, which corresponds to the average receive traffic.

[0046] As shown in Fig. 9(b), a receiving node in a basic configuration in an optical network of the present disclosure includes multiple AWGs 14 that passively route received data from different source nodes, and an interface unit 15. A receiver and a receiving ASIC switch 17 are provided in front of the interface unit 15, and further includes a storage means (memory) 19 that operates in conjunction with the receiving ASIC switch 17. The capacity of the receiving ASIC switch 17 is set to match the average received traffic or to slightly exceed the average received traffic, and the storage means 19 can be used to process traffic that exceeds this switching capacity with a slight delay.

[0047] As explained above, the optical network of the basic configuration of the present disclosure operates in a time slot manner for transmission from a source node to the same destination node in order to limit data reception at the receiving node. The source node must wait for one reception cycle to elapse before it can address the same destination node again. The data reception limit reduces the effective transmission rate and degrades the effective BW of data transmission between node pairs.

[0048] FIG. 11 is a diagram for explaining the deterioration of the effective BW between any pair of nodes due to network division. The horizontal axis represents the number of divisions (number of groups) d of the network, and the vertical axis represents the maximum transmission BW at the receiving node. out 11 shows the received BW normalized by d. As shown in FIG. 11, the received BW decreases with an increase in the number of network divisions d. For example, when the number of divisions d increases from 8 to 32, the received BW decreases from 12.5% ​​to 3.12%. A mechanism is required to compensate for such a decrease in effective BW between node pairs caused by network division. The inventors have proposed two solutions to solve the problem of effective BW decrease in the optical network of the basic configuration: modification to the operation of the source node and modification to the transmission path between any node. One solution allows the source node to reach the desired destination node in one or more time slots, and the other solution increases the effective BW by setting up another optical line between any node pair.

[0049] In this paper, in addition to solving the problem of the reduction in effective BW between node pairs, we propose to modify the time slot configuration or the electrical switch configuration of the receiving node in the optical network of the basic configuration described above. We propose a method to deal with the performance limit of ASIC switches from a new perspective different from the previous solutions. [Maximum incoming bandwidth of the receiving node BW in_max Introduction of a factor F for

[0050] If it is possible to maintain the ability to handle the maximum traffic, with some compromise, the bandwidth BW of the receiving ASIC switch in the network node switch_avg can be set to a fraction F (0 < F < 1) of the maximum receiving bandwidth BW in_max . To cope with the rapidly increasing traffic volume and the limitations of switching technology as described initially, it would be necessary to realistically reduce the switching capacity (bandwidth) in the ASIC switch to an available value. First, the different concepts of bandwidth in each part of the DC network with the basic configuration shown in FIG. 4 will be explained here.

[0051] FIG. 12 is a diagram for explaining different bandwidths of traffic in the DC network. In FIG. 12, the same basic configuration of the DC network of the present disclosure shown in FIG. 4 is shown. Here, various definitions of bandwidth viewed from different perspectives related to the network node are shown.

[0052] (Definition 1) BW out : The total output bandwidth from each node (Definition 2) BW in_max : The maximum incoming bandwidth (to the receiving switch) of each node (Definition 3) BW network : The total bandwidth of the network Referring to FIG. 12, the total bandwidth BW network of the entire network 10 is represented by arrow 22, and the total output bandwidth BW out from each individual node is represented by arrow 23. If the number of nodes in the network 10 is N, the following relationship holds. BW network = BW out × N Equation (1) In FIG. 12, the maximum incoming bandwidth BW in_maxis represented by arrow 24. In the DC network of the present disclosure, as already described in FIGS. 9 and 10, a storage means 19 can be provided in conjunction with the receiving ASIC switch 17. Thereby, the switching capacity of the receiving ASIC switch 17 is set to match the average received traffic or slightly exceed the average received traffic, and the storage means 19 is used to process the traffic exceeding this switching capacity with a slight delay. In FIG. 12, the average bandwidth BW for the receiving ASIC switch switch_avg is represented by arrow 25.

[0053] The inventors considered that it is difficult to flexibly respond to a sudden increase in traffic or a limiting situation of ASIC technology progress by allocating the maximum available bandwidth to the receiving ASIC switch at all times. The bandwidth BW of the receiving switch switch_avg is not only set in advance to match or slightly exceed the average received traffic, but also the inventors came up with the idea of setting it in a wider range as a fraction F (0 < F < 1) of the maximum receiving bandwidth BW in_max . Therefore, in the DC network of the present embodiment, the average bandwidth BW switch_avg for the receiving ASIC switch 17 in FIG. 12 is set to the "reduced bandwidth" where the maximum incoming bandwidth BW in_max is reduced by the coefficient F. In the configuration of the DC network of the present embodiment to cope with the performance limit of the ASIC switch, the following relationship is further defined for the bandwidth of the ASIC switch.

[0054] (Definition 4) BW switch_avg : Reduced bandwidth of the main ASIC switch at the node (Definition 5) Coefficient F: Ratio of the reduced bandwidth BW in_max to the ASIC switch to the maximum incoming bandwidth BW switch_avg to the node From the above definition (5) of the "reduced bandwidth", the following equation is obtained. BW switch_avg = F × BW in_max Equation (2) In the DC network configuration shown in FIG. 4, if it is difficult for the ASIC switch to satisfy the required maximum reception bandwidth BW in_max it may be reasonable to introduce a "reduced bandwidth". By introducing a coefficient F, it is possible to flexibly cope with the bandwidth of an ASIC switch that can be actually used. At the same time, the total network bandwidth BW network can be increased by a factor of the same coefficient F. Furthermore, by restricting the bandwidth of the receiving ASIC switch, it is generally possible to reduce the power consumption of the node. Thus, introducing the coefficient F into the bandwidth of the receiving ASIC switch is beneficial for the scalability of the network.

[0055] FIG. 13 is a diagram for explaining the concept of the reduced bandwidth of the ASIC switch in the DC network of the present embodiment. FIG. 13(a) clearly shows the relationship between the average bandwidth BW switch_avg and the maximum incoming bandwidth BW in_max at the receiving node. As shown in FIG. 13, the average bandwidth BW switch_avg 25 of the receiving ASIC switch 17 is set to a part (Fraction) of the maximum incoming bandwidth BW in_max 24 to the node, and the maximum incoming bandwidth BW in_max 24 is divided by the coefficient F. As described in (Definition 5) above, the coefficient F represents the ratio between the maximum incoming bandwidth BW in_max 24 and the average bandwidth BW switch_avg 25 assigned to the ASIC switch, and is in the range of 0 < F < 1.

[0056] If the number of groups (number of divisions) in the network 10 is d, the number of nodes m belonging to the same group is m = N / d. Also, let T be the reception cycle period during which reception from all node groups is completed.

[0057] The reception cycle during which reception from all node groups is completed is T, and from the configuration of the reception cycle T having a time slot structure corresponding to d groups, the maximum incoming bandwidth BW in_max of each node can be expressed by the following equation. BW in_max =BW out ×m=BW out ×(N / d) Equation (3) Equation (3) is transformed using equations (1) and (2), and from the viewpoint of bandwidth, the number of groups d can also be expressed by the following equation: BW in_max =(BW network / N) × (N / d) d=BW network / BW in_max Formula (4) As shown in Fig. 6 and Fig. 8, one TS time is t TS Then, the reception cycle period T has the following relationship with the number of groups d: T=t TS ×BW network / BW in_max Formula (5) From equations (2) to (5), the following relationship can be obtained: T×BW switch_avg =t TS ×F×BW network Formula (6) In equation (6), the average bandwidth for the ASIC switch, BW switch_avg and the total network bandwidth BW network The relationship between the factor F and BW explains the scalability of the network. network The product (right side) of the receive cycle period T and BW switch_avg For example, the total network bandwidth BW network If fluctuates, the relationship of equation (6) can be maintained by increasing or decreasing the coefficient F in response to this fluctuation. In other words, by varying the set value of the coefficient F, the same reception cycle period T and average bandwidth BW of the reception switch can be maintained. switch_avg The configuration can be maintained.

[0058] Figure 14 shows the average bandwidth (BW) of the receiving ASIC switch. switch_avg 14 is a graph showing the relationship between the average bandwidth BW of the receiving switch in a DC network with a coefficient F.switch_avg (Tbps), the vertical axis is the reception cycle period T (nsec), and the total network bandwidth BW network As a prerequisite for the calculation, we introduce the coefficient F shown in definition (5) for the ASIC switch of the receiving node, and calculate the average bandwidth BW switch_avg The maximum incoming bandwidth to a node is BW in_max The total output bandwidth per node is set to 30% (F = 0.3). out 400Gbps, time slot time t TS 40ns, total network bandwidth BW network The parameters are in the range of 51.2 Tbps to 1.64 Pbps, and six curves are shown.

[0059] From the graph in Figure 14, the average bandwidth of the ASIC switch, BW switch_avg The larger the value, the shorter the reception cycle period T becomes, and conversely, the average bandwidth BW switch_avg The smaller the , the longer the reception cycle period T. The average bandwidth BW of the reception switch that can be processed in real time switch_avg The plot on the left corresponds to the current level of ASIC technology, and the plot on the right corresponds to the expected level of ASIC technology in the future. network It can be seen that the reception cycle period T becomes longer as increases. The reception cycle period T is the period until the next data is received in the time slot method of the DC network disclosed herein, so it is desirable to make it shorter. The graph in FIG. 14 shows the ASIC switch average bandwidth BW switch_avg and total network bandwidth BW network This shows a scalable relationship between

[0060] Next, a more specific node configuration for introducing a "reduced bandwidth" by a factor F into the receiving ASCI switch will be further described. [Handling above average input traffic with extended receive cycle period] Maximum incoming bandwidth of the receiving node BW in_max By introducing a factor F, the average bandwidth of the ASIC switch BW switch_avg If we define the "reduced bandwidth" as the ASIC switch needs to handle the above-average input traffic. The first idea for handling the above-average input traffic is simple and is based on adding one or more time slots that do not actually receive data from any node. This additional time slot is an idle time slot and constitutes an extended receive cycle period.

[0061] Referring again to FIG. 13, FIG. 13(b) shows two concepts of how to handle above-average incoming traffic with an ASIC switch that is bandwidth-limited by a factor F. The first solution, at the top of FIG. 13(b), shows the construction of an extended receive cycle period on the time axis by adding an idle time slot 26 during the receive cycle period T during which no data is received from the node. This idle time slot 26 is used by the receive ASIC switch to temporarily store the average (average bandwidth BW switch_avg ) for switching additional incoming traffic. These idle time slots do not actually receive data from any node, are not addressed, and remain unallocated.

[0062] FIG. 15 is a diagram for explaining the configuration of an extended reception cycle period accompanying the introduction of the coefficient F. The extended reception cycle time slot configuration 31 is shown together with the normal time slot configuration 30 explained in FIG. 6 and FIG. 8. In the example of FIG. 15, the normal time slot configuration 30 is made up of 16 time slots (TS1 to TS16), and each time slot corresponds to one of the 16 groups. In the normal time slot configuration 30, the reception cycle period T33 is TS time x number of groups (16).

[0063] On the other hand, the extended reception cycle period T EXTEND15, the idle time 34 includes a receive cycle period T33 followed by an idle time 34. Although in FIG. 15, the idle time 34 is shown as an additional time slot, the idle time 34 may be one or more time slots. During the idle time 34, no network group is addressed and no data is received. During the idle time 34, the receive ASIC switch can process additional (excess) traffic that could not be processed due to the above average incoming traffic in each time slot. In this solution, the extended receive cycle period T EXTEND The switching process by the ASIC switch is divided into two parts depending on the time in 35. One part is a normal reception cycle period T33 in which the average bandwidth BW is increased with a time limit for each node group. switch_avg The other is a process that is performed during idle time 34, without any time restrictions depending on the node group to which it belongs. This solution does not require any changes to the configuration of the receiving node, and uses the storage means 19 shown in FIG. 9 to store the average bandwidth BW switch_avg More than 100,000 traffic can be handled during idle time 34.

[0064] Therefore, the optical network disclosed herein is an optical network comprising an optical core section having a full mesh network configuration, and a plurality of nodes connected to the optical core section, the plurality of nodes being divided into a plurality of groups, with each group including up to m nodes, each of the plurality of nodes having an ASIC switch that switches electrical signals corresponding to optical signals received from the maximum m nodes and routes them to a plurality of servers, the ASIC switch having a switching capacity corresponding to the average incoming traffic of the plurality of nodes, and during a receiving cycle period including a plurality of time slots, the ASIC switch is addressed by any node in the group to which the source node belongs only in a time slot 33 associated with the group to which the source node belongs, and in one or more idle time slots 34 following the plurality of time slots, no optical signals are received from any of the source nodes, and the optical network can be implemented as processing traffic that exceeds the average incoming traffic.

[0065] Here, i is the index of the time slot, and the input bandwidth in each time slot of one reception cycle period T is BW i The amount of traffic arriving at the ASIC switch changes for each time slot. i and the average bandwidth BW switch_avg The difference between i When adding S idle time slots as the idle time 34, the maximum additional traffic that the receive ASIC switch can handle is limited by the following equation, which averages the additional traffic over one receive cycle period T: TIFF0007626956000001.tif11150

[0066] The average bandwidth already given, BW switch_avg For a value of,,, in order to handle more additional traffic resulting in excess bandwidth, the number of added idle time slots,S,must be increased, as is evident from equation (7).

[0067] When adding idle time slots, the extended receive cycle period T EXTEND This is undesirable because it adds delay to the overall network operation cycle. Next, from another perspective, the average bandwidth BW switch_avg A solution is proposed to handle above-average input traffic when β is the "reduced bandwidth". [Modified configuration with an auxiliary switch added to the main switch] Referring again to the lower diagram in Figure 13(b), we see that an ASIC switch with bandwidth limited by a factor F is unable to handle input traffic that exceeds the average. The The second solution shows the concept of the second solution. In the second solution, a factor F is introduced to reduce the average bandwidth BW of the ASIC switch. switch_avg When the bandwidth is "reduced," the auxiliary switch 27 is provided in addition to the main ASIC switch in order to process input traffic that exceeds the average with the ASIC switch. This is not an approach to the time axis like the first solution, but an approach to modify the switching operation by providing a main switch and an auxiliary switch 27 with different functions. The main switch is a high-speed switch with the performance of the latest technology at the time, while the auxiliary switch 27 has a narrower bandwidth than the main switch and can use a slower switch as described later.

[0068] FIG. 16 is a diagram for explaining the configuration of an ASIC switch modified with the introduction of the coefficient F. FIG. 16 shows a receiving node 40 having an ASIC switch with a modified configuration, which is the same as the configuration of the receiving node of the basic configuration of the DC network according to the present disclosure shown in FIG. 4 and FIG. 9. That is, the front part of the node is provided with AWGs 43, the number of which is m, the number of nodes in one group, and corresponding receivers 44. The node 40 is provided with an auxiliary switch 42 in addition to a main switch 41. The main switch 41 is a high-speed, wideband ASIC switch, like the receiving node of the basic configuration in FIG. 4 and FIG. 9, and has an average bandwidth BW reduced by the coefficient F. switch_avg16. On the other hand, the auxiliary switch 42 is a low-speed switch having a narrow bandwidth set to 1 / q of the bandwidth of the main switch 41. The main switch 41 operates under time restrictions by the time slot method described in Fig. 6 and Fig. 8, but the auxiliary switch 42 can be used without time restrictions throughout the entire reception cycle period T. The main switch 41 and the auxiliary switch 42 are controlled by a control unit of a node not shown in Fig. 16, and the two switches having different bandwidths operate in cooperation with each other.

[0069] Therefore, the optical network disclosed herein is an optical network including an optical core section having a full mesh network configuration, and a plurality of nodes connected to the optical core section, the plurality of nodes being divided into a plurality of groups, with each group including up to m nodes, and each of the plurality of nodes being addressed by any node in the group to which the source node belongs only in the time slot associated with the group to which the source node belongs during a receiving cycle period including a plurality of time slots, and an ASIC switch that switches electrical signals corresponding to the optical signals received from the maximum m nodes and routes them to a plurality of servers, the ASIC switch having a switching capacity corresponding to the average incoming traffic of the plurality of nodes and operating in synchronization with the time slots, and an auxiliary switch having a switching capacity capable of processing traffic exceeding the average incoming traffic and operating regardless of the time slots.

[0070] Since the auxiliary switch 42 can be used throughout the receive cycle period T without time restrictions, the maximum additional traffic that can be handled by the auxiliary switch is given as follows: The input bandwidth in each time slot of one receive cycle period T is expressed as BW i The traffic arriving at the ASIC switch changes for each time slot, and the BW i and the average bandwidth BW switch_avg The difference betweeni Let the number of time slots in the receive cycle period T be d (the number of time slots d also corresponds to the number of groups of nodes). The maximum additional traffic by the ASIC switch in the modified configuration of FIG. 16 is limited by the following equation: From equation (8), the additional traffic ΔBW during the receiving cycle period T i is rather enhanced by the factor d / q on the right hand side, e.g., q=4 and d=16 would achieve an expansion factor of d / q=4. Compared to equation (7) in the first solution where the receive cycle period T is expanded by adding idle time, it is possible to handle input traffic that exceeds the average bandwidth without adding idle time slots. By further modifying the configuration of the ASIC switch that operates in "reduced bandwidth" by a factor F, it is possible to handle input traffic that exceeds the average without incurring the delay problem of the first solution, and even to enhance the amount of excess bandwidth.

[0071] It is also possible to combine the second solution using the modified configuration of the SAIC switch described above with the first solution of extending the receive cycle period T described above. That is, it is possible to provide two switches 41, 42 with an extended receive cycle period by adding an idle time slot as shown in FIG. 15 and with different bandwidths as shown in FIG. 16. In this case, each of the nodes of the network does not receive an optical signal from any source node during one or more idle time slots following the multiple time slots, and processes traffic that exceeds the average incoming traffic. At the same time, the auxiliary switch 42 processes traffic that exceeds the average incoming traffic throughout the entire period of the multiple time slots and the one or more idle time slots. [Minimizing queuing time after switching with auxiliary switches]

[0072] In the second solution using the modified ASIC switch configuration of FIG. 16 described above, the causes of latency (delay) encountered by traffic passing through the auxiliary switch path can be divided into two. The first cause is latency caused by the long physical switching time due to the narrow bandwidth of the auxiliary switch 42. The second cause is latency caused by the extra latency due to queuing until the transmission to the top-of-rack (ToR) switch 45 is completed. In the following, a configuration is proposed that addresses these delays caused by the auxiliary switch 42 by further modifying the second solution of FIG. 16.

[0073] FIG. 17 is a configuration diagram of a node that reduces the delay of the modified ASIC switch. The receiving node 50 having the modified ASIC switch in FIG. 17 has a substantially identical configuration to the receiving node 40 described in FIG. 16, so only the differences will be described. The queue 55 in the auxiliary switch 52 is also included in the receiving node 40 in FIG. 16, and corresponds to the queue in the delay of the second reason described above. In the receiving node 50 in FIG. 17, the auxiliary switch 52 and the ToR switch 57 are connected via a link 56 with a large transmission capacity that can be realized as an optical link. The configuration in FIG. 17 supports a capacity sufficient to transmit data accumulated in one time slot to the target ToR switch, and can reduce the delay in the modified ASIC switch.

[0074] As described above in detail, the optical network disclosed herein can simplify the configuration of nodes in the periphery of a DC network and reduce power consumption. It can solve or at least alleviate the problems that ASIC switches have, and can accommodate larger scale and lower power consumption in optical networks. Furthermore, it can address various limitations of ASIC switches and achieve scalability that can flexibly adapt to network traffic demands and realistic technological advances in ASIC switches. [Industrial Applicability]

[0075] The present invention can be generally applied to optical communication systems.

Claims

1. an optical core unit having a full mesh network configuration; A plurality of nodes connected to the optical core unit, the plurality of nodes being divided into a plurality of groups, with each group including a maximum of m nodes; An optical network comprising: Each of the plurality of nodes an ASIC switch for switching and routing electrical signals corresponding to the optical signals received from the maximum m nodes to a plurality of servers, the ASIC switch having a switching capacity corresponding to an average incoming traffic at each of the plurality of nodes; during a receive cycle period including a plurality of time slots, the source node is addressed by any node in the group to which the source node belongs only in a time slot associated with the group to which the source node belongs among the plurality of nodes; a storage means associated with said ASIC switch for processing traffic in excess of said average incoming traffic during one or more idle time slots following said plurality of time slots in which no optical signal is received from any of said source nodes; An optical network characterized by:

2. an optical core unit having a full mesh network configuration; A plurality of nodes connected to the optical core unit, the plurality of nodes being divided into a plurality of groups, with each group including a maximum of m nodes; An optical network comprising: Each of the plurality of nodes during a receive cycle period including a plurality of time slots, the source node is addressed by any node in the group to which the source node belongs only in a time slot associated with the group to which the source node belongs among the plurality of nodes; An ASIC switch that switches electrical signals corresponding to the optical signals received from the maximum m nodes and routes them to a plurality of servers, the ASIC switch comprising: a main switch having a switching capacity corresponding to an average incoming traffic at each of the plurality of nodes and operating synchronously with the time slots; and an auxiliary switch having a switching capacity capable of handling traffic exceeding said average incoming traffic and operating without regard to said time slot; 1. An optical network comprising:

3. Each of the plurality of nodes a storage means associated with said ASIC switch for receiving, during one or more idle time slots following said plurality of time slots, no optical signal from any of said source nodes and processing traffic in excess of said average incoming traffic; 3. The optical network of claim 2, wherein said auxiliary switch handles traffic in excess of said average incoming traffic throughout the duration of said plurality of time slots and said one or more idle time slots.

4. 4. The optical network of claim 2, wherein each of said plurality of nodes further comprises an additional optical link connecting said auxiliary switch to a top-of-rack switch including said plurality of servers.

5. Each of the plurality of nodes m arrayed waveguide gratings (AWGs), each having a plurality of input ports for receiving corresponding optical signals from the maximum m nodes belonging to a same group, and a wavelength used by one or more source nodes among the maximum m nodes is set to match an operating wavelength of the plurality of input ports; m receivers connected to output multiplexing ports of the AWG; a storage medium for storing and processing traffic exceeding said switching capacity; 5. The optical network of claim 1, further comprising:

6. A maximum incoming bandwidth BW to each of the plurality of nodes in_max As, The ASIC switch is configured to reduce the bandwidth by a factor F (0<F<1) to provide an average bandwidth BW corresponding to the average incoming traffic. switch_avg = F x BW in_max 6. The optical network according to claim 1, wherein:

7. A node operating in an optical network, comprising: The optical network includes: an optical core unit having a full mesh network configuration; A plurality of nodes connected to the optical core unit, the plurality of nodes being divided into a plurality of groups, with each group including a maximum of m nodes; the node being one of the plurality of nodes; The node in question is an ASIC switch for switching and routing electrical signals corresponding to the optical signals received from the maximum m nodes to a plurality of servers, the ASIC switch having a switching capacity corresponding to an average incoming traffic at each of the plurality of nodes; during a receive cycle period including a plurality of time slots, the source node is addressed by any node in the group to which the source node belongs only in a time slot associated with the group to which the source node belongs among the plurality of nodes; a storage means associated with said ASIC switch for processing traffic in excess of said average incoming traffic during one or more idle time slots following said plurality of time slots in which no optical signal is received from any of said source nodes; A node characterized by:

8. A node operating in an optical network, comprising: The optical network includes: an optical core unit having a full mesh network configuration; A plurality of nodes connected to the optical core unit, the plurality of nodes being divided into a plurality of groups, with each group including a maximum of m nodes; the node being one of the plurality of nodes; The node in question is during a receive cycle period including a plurality of time slots, the source node is addressed by any node in the group to which the source node belongs only in a time slot associated with the group to which the source node belongs among the plurality of nodes; an ASIC switch that switches electrical signals corresponding to the optical signals received from the maximum m nodes and routes them to a plurality of servers, the ASIC switch comprising: a main switch having a switching capacity corresponding to an average incoming traffic at each of the plurality of nodes and operating synchronously with the time slots; and an auxiliary switch having a switching capacity capable of handling traffic exceeding said average incoming traffic and operating without regard to said time slot; A node comprising:

Images