JP2008211728A - Awg single hop wdm network backup plane constitution method and network system employing same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accomplish an AWG single hop WDM network backup plane constitution method with which the number of collected nodes of a backup plane is remarkably reduced, costs are reduced and reliability is improved. <P>SOLUTION: An active type coupler CPCA includes a packet collision avoiding mechanism in which only the packet having firstly arrived at the coupler is allowed to pass and other packets are blocked when the coupler is in an idle state on each wavelength channel in the network of a configuration where a plurality of nodes are integrated in a coupler CP and then are connected to an AWG, out of an AWG single hop WDM network targeted to a MAN employing the AWG for a hub node, for a network configuration employing such an active type coupler CPCA as a target, the backup plane for reliability improvement is constituted by paying attention to it that the possibility for a plurality of active type couplers to simultaneously break down is low and by taking into account only the traffic sent from nodes accommodated in the same active type coupler on a normal plane. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an AWG single-hop WDM (Wavelength Division Multiplexing) network for MAN using AWG (Arrayed-Waveguide Grating) as a hub node, and a plurality of nodes are aggregated by a coupler and then connected to the AWG. The network configuration using an active coupler provided with a packet collision avoidance mechanism that allows only packets that first arrive at the coupler to pass and blocks other packets when the coupler is empty on each wavelength channel. Focusing on the fact that multiple active couplers are unlikely to fail at the same time, considering only the traffic sent from the nodes accommodated in the same active coupler on the working surface, to improve reliability Spare surface It relates to a network system using a pre-surface construction method and method for forming.

  Although the transmission capacity of LAN and WAN continues to improve dramatically, SONET / SDH is still the mainstream for the metro area network (MAN) intervening between LAN and WAN. There is a concern that node bottlenecks associated with electrical processing may increase with the increase in the number of nodes. Therefore, MAN is expected to introduce an all-optical network that eliminates packet store-and-forward at nodes and performs cut-through transmission with optical signals. However, since MAN is more strict with respect to cost, it is important to apply an all-optical network that suppresses both facility cost (CAPEX) and control cost (OPEX).

The all-optical network is large and can be classified into a wavelength path network, an optical burst switch network, and an optical packet switch network. From the viewpoint of utilization efficiency of wavelength resources, an optical packet switch network with the smallest switching granularity is superior. However, it is difficult to mount a large-capacity optical memory, and there are many problems to be solved for its realization.
On the other hand, a single-hop WDM network in which all nodes are connected to a single hub node has been studied as a method for realizing an all-optical network in a LAN.

  In recent years, it has been attracting attention as an all-optical network in MAN. Since all packets are directly transmitted via the hub node without passing through other nodes, routing and signaling are unnecessary, and OPEX can be suppressed to an extremely low level. In addition, since a passive optical device is used for the hub node and no replacement process is performed, no optical memory is required. Therefore, it is effective as a method for realizing an all-optical packet network in a LAN or MAN.

  Two types of optical devices used for hub nodes are being studied: star couplers (SC) and arrayed-waveguide gratings (AWG). In the case of the configuration using the SC, since packets arriving at the hub node are broadcast to all nodes, when the transmission band of each wavelength channel is C and the number of multiplexed wavelengths is W, the throughput that can be achieved in the entire network is the maximum. But it becomes CW. Furthermore, access control is required due to the configuration in which the same wavelength channel is shared by a plurality of nodes. For this reason, the use efficiency of wavelength resources is low, and single-hop WDM networks using SC have been studied for LAN.

On the other hand, in AWG, wavelength resources can be spatially reused due to wavelength recursion (packets can be transferred between a plurality of nodes simultaneously using the same wavelength), and the throughput that can be achieved in the entire network is CW ^2. Furthermore, since each node pair can exclusively use the wavelength channel, access control is not necessary. For this reason, a single-hop WDM network (hereinafter referred to as an AWG network) using AWG as a hub node has attracted attention as an all-optical network in MAN with strict cost requirements.

However, in order to enable packet transfer between all nodes in the AWG network, assuming that the number of nodes is N, it is necessary to prepare N transceivers at each node. Therefore, since the total number of transceivers increases in proportion to N ² as N increases, the AWG network has a problem in scalability. For example, when about 200 MAN nodes are considered, the total number of transceivers is 40,000, which is not practical as the number of transceivers installed in one MAN. In order to solve the scalability problem in the AWG network, it is effective to install a coupler and a splitter between the hub node and the node and connect a plurality of nodes together to one AWG port. Hereinafter, this architecture is referred to as an “AWG-CP network”.

However, in this embodiment, packet collision in the coupler becomes a problem when the load is high. In CSMA and CSMA / CD, which are typical MACs of random access networks, throughput deteriorates when the line rate is high. The demand assignment method using fixed time slots requires time slot synchronization, and it is necessary to dynamically secure resources using a control channel.
Such control overhead increases OPEX and is difficult to apply to cost-sensitive LANs and MANs. In order to take advantage of the need for single-hop WDM network signaling, it is desirable that all nodes operate autonomously.
In order to solve this problem, a network architecture in which a packet collision avoidance mechanism is added to a coupler that performs node aggregation in an AWG-CP network has been studied. This network is referred to as an “AWG-CCA network”.

The packet collision avoidance mechanism was originally proposed for the purpose of improving the throughput of a band-sharing LAN using CSMA / CD for hub nodes that handle data using electrical signals. All input ports are always monitored and transmitted. When the resource (shared by all input ports) is free, the first packet that arrives is passed, but all other packets that arrive during transfer are blocked. Even at high loads, at least one packet transfer is guaranteed, so it is possible to avoid a decrease in throughput at high loads that occurs in the ALOHA method or the CSMA / CD method.
An input module for realizing a packet collision avoidance mechanism is installed at each input port of the coupler.

FIG. 1 is a diagram showing a network configuration of an AWG-CCA network and a configuration of an input module added to a coupler.
Packet collision avoidance in the light transmission node can be realized by installing a wavelength selection filter that dynamically switches the passing wavelength according to an instruction from the gate control device at each input port of the CP.
FIG. 1A shows a network configuration, and FIG. 1B shows a coupler input module configuration. Internal traffic, which is AC traffic inside MAN, and external traffic, which is AC traffic between MAN, flows through MAN, but all external traffic flows out to WAN via POP (Point-Of-Presence). Or assume that it flows from the WAN. Therefore, since a large amount of traffic is concentrated on the POP, it is assumed that k optical cores are prepared between the POP and the hub node (AWG). Therefore, each node can transmit / receive a packet to / from the POP using k wavelength channels. Since D CPs (SPs) are installed, the number of AWG ports is D + k.
Packet collision avoidance is realized by installing a wavelength selection filter that dynamically changes the passing wavelength in accordance with an instruction from a gate controller (Gate controller) at each input port of the CP.

FIG. 1B shows a case where a wavelength selective filter having a configuration of a binary optical switch element (SOA or the like) and a wavelength demultiplexing device (Mux, Demux) is used. The optical signal that has arrived at the input module is separated for each wavelength by a wavelength demultiplexer (Demux), a part of which is taken out by a tap, and the presence or absence of the optical signal is detected by a photodetector (Photo detector). .
The gate controller (Gate controller) turns on only the gate of the input module where the packet first arrives without the coupler (CP) transferring the packet for each wavelength, and supports all other input modules. The gate of the wavelength to be turned off is turned off.
Then, when the transfer of the packet is completed, all the gates of the corresponding wavelengths that have been closed are turned on. As a result, in each wavelength channel, only the first packet that arrives when the CP is idle passes through the CP, and packet collision of the packet is completely avoided.

Since the gate opening / closing process operates autonomously in each CP, it is not necessary to exchange any control signals between the CP, the node, and the AWG, and each node can transmit a packet at an arbitrary timing. Packets arriving at the CP when other packets are being forwarded are blocked, so it is necessary to consider the possibility of packet loss. For example, if an ACK is not returned within the timeout period, The transmitting terminal can cope with the packet loss by using a method such as retransmitting the same frame after the back-off time.
For the above technique, refer to Patent Document 1 below.
Noriaki Kamiyama, “AWG Single-Hop WDM Network System, Optimal Coupler Number Calculation Method and Apparatus” (Japanese Patent Laid-Open No. 2006-54648)

JP 2006-54648 A

Since the AWG, coupler, and splitter are all passive optical elements, the AWG network (network in which nodes are aggregated in the AWG) and the AWG-CP network (the AWG through the coupler (CP)) are configured using these. The network in which nodes are aggregated) has a long time interval (MTBF) until a failure occurs, and is very reliable. In contrast, AWG-CCA networks (networks that are aggregated by AWG via couplers (CPCA)) add active devices and optical elements such as gate controllers and SOAs to couplers to implement packet collision avoidance mechanisms. Therefore, there is a concern that reliability may be lowered due to a decrease in MTBF (failure interval) in the node aggregation portion. In MAN, it is necessary to avoid communication interruption due to a failure as much as possible. To that end, it is indispensable that the network has a two-surface configuration in which a spare surface is provided in addition to the current surface.
However, when all network components are simply duplicated, an increase in CAPEX (equipment cost) becomes a problem.

(the purpose)
The purpose of the present invention is to focus on the fact that a plurality of aggregation nodes are unlikely to fail at the same time, and to construct a spare plane considering only the nodes accommodated in the same aggregation node on the active plane. AWG single-hop WDM network spare surface configuration method that greatly reduces the number of aggregated nodes on the spare surface, and is low-cost and highly reliable by sharing the aggregate nodes on the spare surface among nodes accommodated in different aggregate nodes It is to provide a network system using the method.

  The AWG single-hop WDM network spare surface configuration method of the present invention is a network in which a plurality of nodes are aggregated by a coupler in an AWG single-hop WDM network targeted for a MAN using the AWG as a hub node and then connected to the AWG. Targeting network configurations using active couplers with a packet collision avoidance mechanism that passes only the first packet that arrives at the coupler when each coupler is idle on each wavelength channel and blocks other packets Considering that there is a low possibility that multiple active couplers will fail at the same time, considering only the traffic sent from the nodes accommodated in the same active coupler on the working surface, a spare for improving reliability Configure the surface.

Further, in the above-described AWG single-hop WDM network spare plane configuration method, a maximum of M nodes accommodated in each of the D active couplers installed on the active plane are D ′ active nodes installed on the spare plane. Under the premise that the packet is equally accommodated in the type coupler, the FSR (Free Spectral Range) value of the spare surface and the number D ′ of installed active couplers are minimized within a range where the average packet transfer time does not exceed the allowable rate γ.
In the AWG single-hop WDM network spare plane configuration method described above, D ′ active couplers on the spare plane are shared and used between nodes accommodated in different active couplers on the active plane.

  Further, in the above-described AWG single-hop WDM network spare plane configuration method, each node uses a binary switch such as SOA to send a packet only to the active plane when normal, and performs fault detection autonomously. When the working side and the standby side are aggregated using a coupler in the receiving direction, packets are always received from both sides, and when the working side is normal, packets are received via the standby side. ACK is also returned via the working surface.

  According to the present invention, attention is paid to the fact that a plurality of aggregation nodes are unlikely to fail at the same time. By sharing the standby aggregation node among the nodes accommodated in different aggregation nodes, it is possible to greatly reduce the number of aggregation nodes on the spare plane and to configure a low-cost and highly reliable AWG-CCA. .

FIG. 2 is a system configuration diagram of an embodiment according to the AWG single-hop WDM network spare surface configuration method of the present invention.
In FIG. 2, 100 is a device for inputting working surface design information such as the FSR of the working surface and the number of installed active couplers, 101 is a device for designing a spare surface such as the FSR of the working surface and the number of installed active couplers, and 102 is a spare surface. This device outputs the design results of
First, design information on the working surface is input to the working surface design information input device 100, a spare surface design is performed by the spare surface design device 101 based on these pieces of information, and the design result is output from the spare surface design information output device 102. Output.
Next, regarding the AWG single-hop WDM network spare surface configuration method according to the embodiment of the present invention, 1) the spare surface configuration method and 2) the spare surface optimum design method will be described in the following order.

(Preliminary surface configuration method)
Since couplers to which a packet collision avoidance mechanism is added (hereinafter referred to as CPCA (Coupler with Collision Aviationance Mechanism) are installed at geographically distant locations, it can be assumed that the failures occur independently. Although the probability is large compared to the case of the normal coupler (CP), it is still a sufficiently small value, so it is expected that the possibility that two or more CPCAs will fail simultaneously is small.
Assuming that the failure interval (MTBF) of each CPCA follows an exponential distribution with an average x, and the recovery time (MTTR) of a failure CPCA follows an exponential distribution with an average y, the failure time ratio of each CPCA is y / (x + y). . Therefore, the probability P _mb that two or more of the D CPCAs on the working surface fail simultaneously is

となる。
図３は、複数ＣＰＣＡが故障する確率と要求ＭＴＢＦの特性図である。
図３（ａ）では、ＭＴＴＲ（復旧時間）がｙ＝４時間のときの、ＭＴＢＦ（故障間隔）（年）に対する確率Ｐ_ｍｂをＤ（現用面のＣＰＣＡの個数）に対して、ＭＴＴＲ（復旧時間）の三つの値について示している。また、図３（ｂ）では、確率Ｐ_ｍｂ＝０．０１を満たすのに要求されるＭＴＢＦ（故障間隔）をＤ（現用面のＣＰＣＡの個数）に対して、ＭＴＴＲ（復旧時間）の三つの値について示している。
図３（ａ）の図からは、故障間隔（ＭＴＢＦ）＝２年の場合、ＣＰＣＡの個数が４のときに、２つ以上が同時に故障する確率Ｐ_ｍｂは１０^−４であるが、ＣＰＣＡの個数が３２のときには、約Ｐ_ｍｂ＝１０^−２．５となることが確認できる。また、図３（ｂ）の図からは、ＣＰＣＡの個数が４５の場合、復旧時間（ＭＴＴＲ）がｙ＝２年のときに、故障間隔（ＭＴＢＦ）は１年であるが、復旧時間（ＭＴＴＲ）がｙ＝４年のときには、故障間隔（ＭＴＢＦ）は２年であり、また復旧時間（ＭＴＴＲ）がｙ＝６年のときには、故障間隔（ＭＴＢＦ）は３年となることが確認できる。
例えば、ルータのＭＴＢＦ（故障間隔）が数年から十数年であることを考えると、２つ以上のＣＰＣＡが同時に故障する可能性（確率Ｐ_ｍｂ）は十分に小さいことが確認できる。

It becomes.
FIG. 3 is a characteristic diagram of the probability that a plurality of CPCAs fail and the required MTBF.
In FIG. 3A, the probability P _mb for MTBF (failure interval) (year) when MTTR (recovery time) is y = 4 hours is MTTR (recovery) with respect to D (number of CPCAs on the active surface). It shows three values (time). Further, in FIG. 3B, the MTBF (failure interval) required to satisfy the probability P _mb = 0.01 is set to three (MTTR (recovery time)) with respect to D (number of CPCAs on the working surface). The value is shown.
From the diagram of FIG. 3 (a), in the case of failure interval (MTBF) = 2 years, when the number of CPCA is 4, the probability P _mb of two or more failures simultaneously is 10 ⁻⁴ , When the number is 32, it can be confirmed that approximately P _mb = 10 ^−2.5 . 3B, when the number of CPCA is 45 and the recovery time (MTTR) is y = 2 years, the failure interval (MTBF) is 1 year, but the recovery time (MTTR) ) Is y = 4 years, the failure interval (MTBF) is 2 years, and when the recovery time (MTTR) is y = 6 years, the failure interval (MTBF) is 3 years.
For example, considering that the MTBF (failure interval) of a router is several years to several tens of years, it can be confirmed that the possibility (probability P _mb ) of two or more CPCAs to fail simultaneously is sufficiently small.

Each node is connected to both the working surface and the spare surface, and the following control is assumed.
a) Each node uses a binary switch such as SOA to send a packet only to the active side when it is normal.
b) Each node autonomously detects a failure, and autonomously switches to a backup plane when a failure is detected.
c) Each node aggregates the active and standby planes using a coupler in the reception direction, and always receives packets from both sides.
d) A node having a normal working surface returns an ACK via the working surface even when a packet is received via the backup surface.
In this case, when the CPCA on the active plane fails, the traffic flowing into the backup plane is limited to that sent by the node accommodated in the failed CPCA.

FIG. 4 is a configuration diagram of a spare surface according to an embodiment of the present invention.
Referring to FIGS. 3 (a) and 3 (b), it is confirmed that it is sufficient to install a spare surface only on M nodes accommodated in a single CPCA instead of all nodes when designing the spare surface. it can.
That is, based on the assumption that a plurality of CPCAs do not fail simultaneously, it can be logically considered that the spare surface constitutes D independent surfaces. Therefore, in the present embodiment, a spare surface configuration method shown in FIG. 4 is proposed.

As shown in FIG. 4, D CPCA (coupler) and SP (splitter) are installed on the working plane (normal plane), and N nodes are equally accommodated in these CPCA. When M is the maximum number of nodes accommodated in a single CPCA, M = [N / D], and the number of nodes accommodated in each CPCA is M or M−1.
On the other hand, D ′ CPCA and SP are installed on the backup plane, and a maximum of M nodes accommodated in the same CPCA on the current plane are equally accommodated in these D ′ CPCA. . If M ′ is the maximum number of logical nodes accommodated in a single CPCA on the spare side, M ′ = [M / D ′], and the logical number of nodes accommodated in each CPCA is M ′ or M′−. 1

Since the D node groups on the working plane share these D ′ CPCAs and SPs on the spare side, physically a maximum of DM ′ nodes are accommodated in a single CPCA on the spare side.
FIG. 4 shows a case where D = 3 and D ′ = 2. By sharing D ′ pieces of CPCA and SP on the spare surface among nodes accommodated in different CPCAs on the active surface, the cost of the spare surface can be greatly reduced.
By the way, since the external traffic does not perform node aggregation by the coupler, the spare surface is not considered here, but in preparation for the failure of the hub node (AWG), the POP (Point Of Presence) is also accommodated in the spare AWG. You can also think about it. Thereby, the reliability of a hub node (AWG) is also improved.

  F wavelength channels are provided between each CPCA and SP due to the wavelength recursion of AWG. When a node transmits a packet, it is necessary to select and use a wavelength channel corresponding to the SP in which the target node is accommodated. In this embodiment, since the combination of nodes accommodated in each CPCA and SP is different between the active surface and the spare surface, the node that detects the failure and switches to the spare surface simultaneously changes the wavelength to be used.

(Preliminary surface optimum design method)
In consideration of static traffic, it is assumed that the ratio of the internal traffic in the total traffic amount V is r, the allowable value of the average packet transfer delay is γ, and the traffic exchange is all uniform and unicast. Each node sends a packet of fixed length P according to the Poisson process, and sets a round trip time between hosts on which retransmission control is performed to a constant value R. The transfer delay time at success tau ₁ packet transfer, the tau ₂ is the average retransmission latency, assuming γ> τ _1.
In the active plane, the spare plane may be designed for a maximum of M nodes accommodated in a single CPCA. Assuming uniform traffic, the amount of internal traffic d transmitted from a certain node i is d = rV / N. The probability that the target node is accommodated in a certain SP is given by N / D ′ = NM ′ / M, and F independent wavelength channels are prepared between each CPCA and SP. The amount of traffic v applied to the wavelength channel is

となる。
パケット衝突回避機構を用いた場合の平均パケット転送時間Ｔは、Ｔ＝(Ｃτ_１＋τ_２ｖ)／(Ｃ−ｖ)で与えられ、Ｔ≦γを満たす必要があることから、

It becomes.
The average packet transfer time T when the packet collision avoidance mechanism is used is given by T = (Cτ ₁ + τ ₂ v) / (C−v), and T ≦ γ needs to be satisfied.

を満たす必要がある。
任意のＦに対して、上式（３）を満足するＭ’(Ｆ)が以下のように得られる。

It is necessary to satisfy.
For any F, M ′ (F) that satisfies the above equation (3) is obtained as follows.

Ｍ’(Ｆ)≧１を満たす必要があることから、Ｆの最適値Ｆ_０’が、

Since M ′ (F) ≧ 1 must be satisfied, the optimum value F ₀ ′ of F is

と得られる。
得られたＦ_０’を用いてＭ’＝Ｍ’(Ｆ０’)よりＭ’を設計することができ、 Ｄ’＝「Ｍ／Ｍ’］よりＤ’が設計できる。

And obtained.
M ′ can be designed from M ′ = M ′ (F0 ′) using the obtained F ₀ ′, and D ′ can be designed from D ′ = “M / M ′].

(Evaluation conditions)
Hereinafter, the numerical evaluation results of the present invention will be described.
C = 10 Gbps, P = 1500 bytes, r = 0.5, and γ = 1 × 10 ⁻³ seconds. Further, the distance between end terminals was 10 km, and the average backoff time during packet retransmission was 10 packet hours. In this case, τ1 = 1.010 × 10 ⁻⁴ seconds and τ ₂ = 1.2 × 10 ⁻⁵ seconds. The total traffic amount V was set as a constant amount, and evaluation was performed by changing the number N of nodes (the transmission / reception traffic amount per node changed as N increased or decreased). Two values of 1 Tbps and 10 Tbps were considered as the value of the total traffic volume V (for example, if the average throughput of each user is 10 Mbps, the former is equivalent to 100,000 users and the latter is equivalent to 1 million users, respectively) ).

(Equipment on the working surface)
FIG. 5 is a characteristic diagram of the number of installed active couplers on the working surface.
FIG. 5 shows the number of installed couplers D of the AWG-CP network and the number of installed CPCAs D of the AWG-CCA network, respectively. FIG. 5A shows a case where the total traffic amount is 1 Tps (the number of 100,000 users when the average throughput of each user is 10 Mbps), and when the number of nodes is N = 10, both the AWG-CP and the AWG-CCA are couplers. The number is 5, but when the number of nodes is N = 100 (1 million users), the number of couplers is 20 in AWG-CP, whereas the number of couplers is 8 in AWG-CCA. So 8 is enough. FIG. 5B shows a case where the total traffic volume is 10 Tbps (1 million users when the average throughput of each user is 10 Mbps), and when the number of nodes is N = 10, both AWG-CP and AWG-CCA are couplers. Although the number is about seven, when the number of nodes is N = 100 (one million users), the number of AWG-CP couplers is about 55, whereas the number of AWG-CCA couplers is Only 30 pieces are enough.

FIG. 6 is an FSR characteristic diagram of the working surface.
By introducing a packet collision avoidance mechanism in the coupler in the AWG-CCA network, more nodes can be aggregated in the coupler than in the AWG-CP network.
In fact, in the region where N is large, it can be confirmed that the AWG-CCA network halves D compared to the AWG-CP network. On the other hand, in a region where N is small, the amount of AC traffic between node pairs is large, and there is no difference in D between the two networks. The CCA network greatly reduces the FSR, and the effect of introducing the packet collision avoidance mechanism into the coupler can be confirmed.

(Equipment on spare side)
FIG. 7 is a characteristic diagram of the number of active couplers installed on the spare surface.
In FIG. 7, the number D ′ of CPCAs installed on the spare surface is shown for the present invention and when all the CPCAs are simply duplicated (Naive backup plane).
FIG. 5A shows a case where the total traffic amount is 1 Tbps (the number of 100,000 users when the average throughput of each user is 10 Mbps), and when the number of nodes is N = 10, D ′ = 5 in a simple duplex case. In contrast, in the case of the present invention (Proposed backed up nodes), D ′ = 2. Further, when the number of nodes is N = 100 (1 million users), D ′ = 8 in the case of simply duplexing, whereas D ′ = 3 in the case of the present invention.
FIG. 5B shows a case where the total traffic amount is 10 Tbps (one million users when the average throughput of each user is 10 Mbps), and when the number of nodes is N = 10, D ′ = 5 in the case of simple duplexing. In contrast, in the case of the present invention, D ′ = 2. Further, when the number of nodes is N = 100, D ′ = 32 in the case of simply duplexing, whereas D ′ = 3 in the case of the present invention.
As is clear from this, when N is small, the CPCA number D on the working surface is small, and the effect of deleting the CPCA number on the spare surface is small according to the present invention. However, in the region where N is larger than about 10, it can be confirmed that the CPCA number reduction effect of the present invention is significant.

FIG. 8 is a characteristic diagram of the FSR of the working surface and the spare surface.
When N is small, as shown in FIG. 7, the difference between D ′ and D is small and the effect of reducing the number of CPCA on the spare surface is not seen, but the present invention can suppress the FSR of the spare surface. Can be confirmed. Further, when N is very large, the amount of AC traffic between each pair of nodes becomes small, and FSR = 1 for both the working surface and the spare surface. However, when N is medium, the effect of reducing the number of CPCAs on the spare surface of the present invention is large, and as a result, the FSR of the spare surface exceeds the FSR of the working surface. The number of transceivers required at each node for transmission / reception of the spare side packet is D′ F ₀ ′.

FIG. 9 is a characteristic diagram of the required number of transceivers on the active side and the spare side.
As seen in FIG. 8, in the region where N is medium, the FSR of the spare surface increases from the FSR of the active surface. However, as seen in FIG. It can be confirmed that the total number of transceivers on the spare surface determined by the product of these is larger than that and is reduced more than the required number on the active surface. Therefore, it can be confirmed that it is not necessary to newly install a transceiver in the node in order to construct a spare surface in all N regions.

Next, CAPEX (equipment cost) generated by configuring the spare surface will be compared.
Since the transmission distance is short in the network scale of the metro area, the cost related to the transmission of the optical fiber, the optical amplifier, etc. is small compared to the cost of the node. Further, the necessary number of CP (coupler), SP (splitter), and gate control device is small compared with the necessary number of other devices, and the cost itself of these devices is small, so it is not considered.
On the other hand, it is assumed that an EDFA optical amplifier for amplifying an attenuated optical signal is installed in all CPCA and SP. As cost components of the spare surface, a wavelength demultiplexer, an EDFA optical amplifier, an SOA, a tap, and a photodetector are considered. SOA, tap, a set per relative cost summarizing three optical devices of the photodetector _{C others,} the relative cost of the transceiver _{C tr,} the relative cost of a wavelength demultiplexer per wavelength _C dm, EDFA optical amplifier _Let C _{edfa be} the relative cost of.
In consideration of the current price, C _tr = 10, C _dm = 1, C _edfa , = 4, and C _others = 2 are set.

FIG. 10 is a characteristic diagram of the relative cost of the spare surface.
FIG. 10 shows a comparison of the relative cost of the spare surface when the present invention simply duplicates all CPCA. As can be seen in FIG. 7, in the small region of N, the number of installed CPCAs D ′ of the spare surface does not change as N increases, but as shown in FIG. 8, since F ₀ ′ decreases, the spare surface The number of multiplexed wavelengths per optical fiber W ′ = D′ F ₀ ′ decreases, and the relative cost of the spare surface decreases.
On the other hand, in the region where N is large, D ′ slightly increases as N increases, and the number of logical accommodation nodes M on the spare surface increases, so the relative cost of the spare surface increases.
However, it is confirmed that the cost of the spare surface can be reduced by about one digit or more by using the present invention in all N regions. In particular, when V and N are large, it is possible to reduce the spare surface cost by two digits or more.

As described above, in the present invention, 1) a spare surface configuration method for an AWG single-hop WDM network configured such that a plurality of nodes are aggregated by a coupler and connected to an AWG that is a hub node. When an active coupler is provided with a packet collision avoidance mechanism that passes only the packet that first arrived at the coupler when the coupler is idle on the channel and blocks other packets, it is installed on the working surface. A maximum of M nodes accommodated in each of the active couplers are equally accommodated in D ′ active couplers installed on the spare surface so that the average packet transfer time does not exceed the allowable value γ. The FSR value of the spare surface and the number D ′ of installed active couplers are minimized.
With the above configuration, it is possible to greatly reduce the number of aggregated nodes on the spare surface and form a highly reliable AWG-CCA network at low cost.
2) Even if the number of active couplers D ′ is not minimized, it is sufficient to configure a spare plane only for traffic transmitted from nodes accommodated in the same active coupler on the active plane. High reliability can be provided at low cost. This means that if the spare surface is configured only for the traffic to be sent out, the number of active couplers described above means that “the spare surface is configured focusing on only the traffic that flows to the spare surface with a single failure”. This has the same effect as when D ′ is minimized.
In this case, a spare plane is formed only for traffic transmitted from a node accommodated in the same active coupler on the active plane, and the spare plane network is maintained for a certain period. Since the continuous period is designed assuming static traffic, it is designed by estimating the amount of traffic excessively so that an expected periodic fluctuation such as a week can be sufficiently accommodated. In that sense, it is assumed that the expansion process will be performed in a period such as one year, so the period during which the same configuration is maintained is expected to be about one year.

It is a block diagram of an AWG-CCA network. 1 is a system configuration diagram of a network according to an embodiment of the present invention. FIG. 11 is a characteristic diagram of a probability that a plurality of CPCAs fail and a required MTBF. It is a block diagram of the reserve surface of this invention. It is a characteristic view of the number of active couplers installed on the working surface. It is a characteristic view of FSR of the working surface. It is a characteristic view of the number of active couplers installed on the spare surface. It is a characteristic figure of FSR of an active surface and a reserve surface. It is a characteristic figure of the required number of transceivers of an active surface and a spare surface. It is a characteristic view of the relative cost of a reserve surface.

Explanation of symbols

100: Active surface design information input device 101: Preliminary surface design device 102: Preliminary surface design information output device POP: Point Of Presence
AWG: Arrayed-waveguide grating
CP: Coupler SP: splitter T ₁₁ through T _DM: Transmitter R ₁₁ to R _DM: Receiver Node: Node SOA: 2 binary optical switching element the Tap: Tap Phto detector: photodetector Demux: wavelength separation device Mux: Wavelength demultiplexer gate controller: gate control device

Claims (8)

A standby plane configuration method for an AWG single-hop WDM network configured to aggregate a plurality of nodes with a coupler and connect to a hub node AWG,
Using an active coupler with a packet collision avoidance mechanism that passes only packets that first arrive at the coupler when each coupler is idle on each wavelength channel and blocks other packets,
An AWG single-hop WDM network spare plane configuration method, wherein a spare plane is configured only for traffic transmitted from a node accommodated in the same active coupler on the active plane. The AWG single-hop WDM network spare surface configuration method according to claim 1,
A maximum of M nodes respectively accommodated in the D active couplers installed on the active plane are equally accommodated in the D ′ active couplers installed on the spare plane;
An AWG single-hop WDM network spare plane configuration method characterized in that the FSR value of the spare plane and the number of installed active couplers D ′ are minimized so long as the average packet transfer time does not exceed the allowable value γ. The AWG single-hop WDM network spare surface configuration method according to claim 1,
The AWG single-hop WDM network spare plane configuration method characterized in that D ′ active couplers on the spare plane are shared between nodes accommodated in different active couplers on the active plane. When the coupler is idle on each wavelength channel, only the first packet arriving at the coupler is allowed to pass, and an active coupler provided with a packet collision avoidance mechanism that blocks other packets is used. A working network configured to be aggregated by the D couplers or splitters and connected to an AWG as a hub node;
An AWG single-hop WDM network system, comprising: a backup plane network configured only for traffic transmitted from nodes accommodated in the same active coupler in the active plane network. In the AWG single hop WDM network system according to claim 4,
Each of the nodes sends out a packet only to the working surface network at the normal time by the binary switch element,
Failure detection is always performed, and when a failure is detected, the working network is switched to the backup network, packet reception is always performed on both the working network and the standby network, and if the working network is normal, the backup network is routed. When the packet is received, an AWG single-hop WDM network system returns an ACK via the active network. In the AWG single hop WDM network system according to claim 4,
An AWG single-hop WDM network system in which D node groups of the active network share a splitter and an active coupler provided with D 'packet collision avoidance mechanisms in the backup network. In the AWG single hop WDM network system according to claim 4,
An AWG single-hop WDM, in which an active coupler and a splitter provided with a D ′ packet collision avoidance mechanism of the backup plane network are shared among nodes accommodated in different active couplers in the active plane network. Network system. In the AWG single hop WDM network system according to claim 4,
An AWG single-hop WDM network system, wherein a POP that transmits and receives external traffic is accommodated in the AWG of the backup network.

Images