JP3366468B2 - Optical switch path selection control method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a route selection control method for an optical switch, and more particularly, it is capable of comprehensively handling a wide variety of media such as images, voices, data, etc. having different required delay qualities. The present invention relates to a route selection control method for a media optical switch.

[0002]

2. Description of the Related Art Conventionally, as a technique in such a field,
For example, IEEE TRANSACTIONS ON
COMMUNICATIONS. VOL. COM-3
5, NO. 5. May 1987 "Routing
in the Manhattan Street N
There was one disclosed in “work”.

Hereinafter, the MSN (Manhattan)
The structure of the Street Network) and its characteristics will be described. FIG. 7 is a diagram showing the structure of such a conventional 6 × 6 MSN. As shown in this figure, the structure is grid-like, and the rows and columns are numbered sequentially from 0 to m−1 (in the case of the m × m structure). For the absolute address of the node, the row and column numbers are used as they are to simplify the routing rule.

Each node has two incoming links and two incoming links.
I have an outbound link for a book. The direction of the link is row,
Alternate for each column. That is, if the links in the odd numbered rows are facing in one direction, the links in the even numbered rows are facing in the opposite direction. Also, the links are circular. The features of the structure of this MSN are shown below.

(1) The same routing rule can be used in all nodes. (2) Addition of nodes is easy and resistant to failures. Since the network has a ring structure, the routing (routing control) is determined only by the relative position (relative address) of the sending node to the destination. That is, since the relative address is used for the routing, the same routing rule can be applied to all nodes regardless of the absolute address.

When adding a node or changing a link, only a part for adding a new node is required. However, in order to maintain a completely uniform network structure, it is necessary to add an even number of nodes for each row and column. Moreover, when a node is added, it may be necessary to change the address of an existing row or column. When a node or link fails, you have to route it so that it doesn't go through. Since the MSN has a lattice type mesh structure, it has many detours and can flexibly cope with a failure.

An outline of the MSN system will be described below. This MSN is a slot periodic system,
The packet is fixed to a certain size. Hereinafter referred to as a cell. The routing decision is made for each cell at each node. Since two links go in and out of one node, cells arriving simultaneously from two inbound links may go out to the same outbound link. In that case, collision is avoided by forcibly sending (deflecting) one cell to the other outgoing line link. In this method, for example, even if the output buffer is 0, no cell is lost.

However, the probability of deflection can be reduced by providing the output line buffer. External (host)
Cells from are sent only when the outgoing link is free. If the cell is deflected, it will take a longer path to the destination than the shortest path, which will prevent new cells from entering the network. Therefore, there is a correlation between the outbound buffer size and the network throughput, and the larger the buffer size,
Increases throughput.

In MSN, even when deflected, the length of the path to the destination node increases by at most 4 links, once only. In addition, there are a large number of nodes that provide the same length route to the destination for both of the two outgoing links, and the possibility of going in a different direction from the destination even when the outgoing links conflict with each other is reduced. Next, a basic routing method will be described.

As the routing method, distributed routing rules, random routing, etc. have been proposed. Here, two routing methods in the distributed routing rule will be described. (1) Rule 1 The routing is determined for each cell in each node. That is, if the cell arrives at the node and the destination is not its own, the node itself selects a link that is expected to arrive at the destination in the shortest time, and transmits the cell to the link.
To determine the link, first find the relative address of the node to the destination.

Since the MSN has a ring structure, any node can be considered to be at the center of the network.
Therefore, first, the destination node is considered to be at the center of the network, and its relative address is set to (0,0). The transmitting node of the absolute address (r _cu , c _cu ) obtains the relative address (r, c) from the position relative to the destination absolute address (r _de , c _de ). FIG. 8 shows the state of the relative address for a certain destination.

Since the destination is considered as the center of the network and the relative address (r, c) is obtained, its range is limited as follows when the size of the network is m × n. -M / 2 <r≤m / 2 -n / 2 <c≤n / 2 Further, it is known from the relative address range in which quadrant the transmitting node is located with respect to its destination.

[0013]

[Table 1]

Once the quadrant is known, it determines the direction of routing. That is the direction of the arrow shown in FIG. From the link direction of the transmission node and the output link of the cell from FIG. 9, the following rules are determined. (1) If there is only one link in the direction indicated by the arrow, select it. (2) If there is no link in the direction indicated by the arrow, or if there are two links, select either one.

To satisfy this rule, the adjacent node in the row (r _nxt, c) and the adjacent node in the column (r, c)
_nxt ) needs to be calculated. As a result, it can be known whether or not the transmitting node is located at r ₁ to r ₄ and c _{1 to} c ₄ shown in FIG. When the quadrant in which the transmitting node is located and the relative address are as shown in Table 2, the transmitting node is located in these rows and columns. These lines,
The columns need to determine unusual routing directions, as shown in the figure.

If routing is performed according to this rule, the destination arrives at the shortest number of hops.

[0017]

[Table 2]

(2) Rule 2 This method is basically the same as Rule 1. However, in the route selection, the direction shown in FIG. 10 is selected instead of that shown in FIG. The difference between the two is the presence / absence of special routing rules in r ₁ , c ₁ , r ₃ , and c ₃ . Since Rule 2 does not have this, there is an advantage that the calculation amount is smaller than that of Rule 1. However, conversely, when the transmitting nodes are located in these rows and columns, there is a case where the route travels slightly longer than the shortest route. However, this will only happen if there is no link at the sending node pointing in the direction of the arrow shown. Further, since these rows and columns are located farthest from the destination, their influence is relatively small in terms of average path length.

[0019]

However, when the above-mentioned conventional MSN is directly applied as an optical switch for handling multimedia information, there are the following problems. That is, all cells are switched with the same level of priority. As the input traffic increases, the probability of being deflected at each node increases in the conventional MSN, and the delay time until it goes out of the switch becomes long. When a switch process is performed without priority, in a service in which real-time property is important, the delay in this switch may significantly affect its quality.

Although it is possible to deal with the contention of the output link by temporarily storing the information in the buffer without performing the deflection, the optical memory and the optical logic circuit are not in the practical stage, and therefore the large-scale optical buffer and the complicated circuit are not available. It is difficult to perform proper light control. SUMMARY OF THE INVENTION It is an object of the present invention to eliminate the above-mentioned problems and to provide a route selection control method for an optical switch that can perform high-speed switching processing with a simple configuration.

[0021]

In order to achieve the above-mentioned object, the present invention has (1) two incoming links and two outgoing links.
In both cases, the direction of the link alternates between rows and columns.
One the link has an annular structure, slot period system
In a method for controlling route selection of an optical switch , which comprises a network system including the above-mentioned two incoming links and two outgoing links as a basic unit, and the nodes are configured in an m × m grid pattern, Due to the selection of the outgoing link in advance to the cells transmitted on the link , its delay to each cell
We also added a bit that represents the priority on demand,
And when there is an outgoing link conflict at the node
Compare the priorities of the two and select the cell with the highest priority as desired.
Send to the outgoing link, lower priority cells to another link
If there is a conflict between the step of assigning a priority for sending and deflecting and the outgoing link due to the cells transmitted to the incoming link,
Outputting to the outgoing line link according to the priority.

(2) In the optical switch path selection control method described in (1) above, when the outgoing line links conflict with each other, a cell having a lower priority is deflected and the deflection data is recorded in the cell. It is the one. (3) In the optical switch path selection control method according to (1), when the priorities are the same, cells are randomly output to the outgoing line link and cells that are not output to the desired outgoing line link. The deflection data is recorded in the above.

(4) In the optical switch path selection control method according to (2) or (3), the deflection data is counted, and if the counter value exceeds a certain threshold value, the priority is raised. It was done.

[0024]

According to the present invention, the cell having the highest priority with respect to the delay takes the destination route through the desired route as long as the node on the way does not conflict with the cell having the same priority for the outgoing link. It can reach the node and send it out of the switch, and can perform high-speed switch processing with a simple configuration.

[0025]

Embodiments of the present invention will be sequentially described below with reference to the drawings. 1 is a diagram showing a basic configuration of a multimedia optical switch showing an embodiment of the present invention, FIG. 2 is a diagram showing an example of a node of the optical switch, and FIG. 3 is an O / E conversion in the node of FIG. It is a figure which shows the internal structure of the node at the time of performing switching by.

It is said that the maximum number of frequency channels that can be used in light is about 1000 channels. In this optical switch, as shown in FIG. 1, channels are alternately allocated to each of the MSN row and column links.
That is, since the same frequency is repeatedly used, only four channels are used in the entire switch. Here, this optical switch is a frequency alternating allocation type MSN (Alt
eternity Frequency Assignment
nt MSN) optical switch, hereinafter referred to as AFAMSN
Abbreviated as optical switch. If the number of rows and columns of the grid type is an even number when performing such channel assignment,
The allocation rule does not change even if the size of the switch is changed.

Since MSN is basically an even number in rows and columns, this allocation method is considered to be suitable. The frequency channels handled by one node are four, two reception channels and two transmission channels. For example, if the receiving channels are f ₀ and f ₁ , the transmitting channels are f ₂ and f
It becomes ₃ . A cell entering from an incoming link is switched to a desired channel and transmitted according to the selection of the outgoing link. Only one channel is used on each link, and since they are unidirectional communication, channel collision does not occur.

As shown in FIG. 3, when an example of the internal structure of the node is shown, the frequency f ₀ input to the input line (row) 1,
When the frequency f ₃ is applied to the input line (column) 2, the frequency f ₀ is the optical / electrical converter (O / E) 3 and the frequency f _{3 is}
Are optical / electrical converters (O / E) 4, each of which is converted into an electric signal, input to the routing switch 5, controlled by the routing switch 5, and output as an electric signal. The frequency f ₁ is converted into an optical signal by the E / O) 6 and the frequency f ₁ is converted into an optical signal by the output line (row) 8 and by the electric / optical converter (E / O) 7.
₂ is output to each output line (column) 9. Also,
Reference numerals 10 and 11 are connection lines to the outside of the network to which terminals and the like connected to the routing switch 5 are connected.

Here, the processing for detecting cell contention is performed by the routing switch 5. When cell contention occurs, the priority bit of the cell is determined, the priority cell is sent to the target output side, E / O converted, and output. Output to the row (column) or column (row) of. Routing switch 5 for non-priority cells
After changing the output destination with, the data is E / O converted and output to the output row (column) or column (row). Details will be described later in a flowchart.

A routing algorithm processing device is provided in each routing switch 5. here,
Explaining the role of the frequency channel, here, multiplexing of information is realized by frequency. In this embodiment, both the input line (2 lines) and the output line (2 lines) are 1
The degree of multiplicity on the book is two, and in consideration of the switch-through function for realizing hierarchization, it is two.

Next, the switching algorithm of this AFAMSN optical switch will be described. The basic system configuration of this switch and the rule for determining the routing direction are the same as those of the conventionally proposed MSN. However, a switching algorithm is used in which cells that are sensitive to delay are preferentially sent to the outside of the switch by performing switching according to the delay requirements of cells that arrive from the outside.

To explain the outline, first, each cell is added with a bit representing a priority corresponding to its delay request. Based on this, when there is a conflict of outgoing links at a node, the priorities of the two are compared, the cell with the higher priority is sent to the outgoing link as desired, and the cell with the lower priority is on another link. Send to. That is, it is deflected. When the output line buffer is provided, the output line buffer 0 does not have to be deflected, but the output line buffer 0 is considered here. If the two have the same priority, one is randomly selected and deflected.

In this switching algorithm,
If the routing direction is determined using Rule 1 described above, the cell with the highest priority is sent out of the switch via the shortest path unless it conflicts with other cells of the same priority. become. Hereinafter, all possible states in each node will be described.

Regarding the outgoing line link selection (rule 1 or rule 2), cells are classified into two types as follows. Type 0 ... Cell where both links are destined to destination, or neither link is destined to destination Type 1 ... Cell where only one link is destined to output Destination cell 5 and the combination of
Street conditions can occur.

(1) When only one cell is output in that slot Type 0 ... Regardless of priority, either link is arbitrarily selected and transmitted. Type 1 ... Send to the determined link regardless of priority. (2) When there are two cells output in that slot Type 0 and type 0 ... Regardless of priority, either link is arbitrarily selected and transmitted so that contention between the two does not occur.

Type 0 and Type 1 ... Regardless of priority,
The outgoing link of the type 0 cell is selected and transmitted so that the type 1 cell does not conflict with the desired outgoing link. Type 1 and Type 1 ... When both separate outgoing links are selected, they are sent to each link. If the same outgoing link is selected, the priorities of the two are compared. If one has a higher priority, the cell is sent to the selected outgoing link and the other lower priority cell is sent to another outgoing line. If the two have the same priority, one of them is arbitrarily selected and deflected.

Cells arriving from the outside are transmitted only when the outgoing link is free. In other words, when a node arrives at a certain node, if there is no relay cell to be transmitted at the next timing, it is the above (1), and if there is one, it is the above (2).
Sent according to the selection algorithm of. When two exist, it will wait until the output becomes empty. Next, the node increasing / decreasing function will be described.

The following are required functions in the node. (1) A routing function that determines outgoing links for each cell. (2) A buffering function for storing cells arriving from the outside. (3) A frequency conversion function that switches the frequency for each cell. As for routing, one node is
This is done by independently determining the outgoing link for each cell. When the outgoing link is determined, the cell is converted into a frequency corresponding to each link and sent out. In order to switch the light as it is, a high-speed frequency converter is required.

When the cell arrives from the outside, if the outgoing line link is not empty, it must be stored in the buffer and wait until the link becomes empty. If the control of this buffer is performed by arrival sequence processing (FIFO), there is a possibility that it can be realized by an optical delay line. Frequency conversion is also required when outputting from this buffer to the outgoing link. FIG. 4 is a flow chart of data according to the route selection control method of the optical switch of the present invention.

As shown in this figure, the incoming links L _i1 , L
Receive cells C _e1 and C _e2 from _i2 (step S1)
Then, it is determined whether or not the outgoing line link L ₀₁ is in conflict (step S2). As a result, in the case of NO, the cells C _e1 and C _e2 are respectively connected to the desired outgoing line links (step S3). YE
In the case of S, the priorities of the cells C _e1 and C _e2 are compared (step S4), and if the priorities are the same, the cells are randomly connected to the outgoing link (step S5) and the cells are deflected. Stores the deflection history data (step S6). Furthermore, the process returns to step S4, C _e1 <For C _e2, determines whether the cell C _e2 (step S7) to, in the case of the cell C _e2, outgoing desired in favor of the cell C _e2 Connect to the link (step S8). If cell C _e1, and stores the deflection historical data to deflection cell C _e1 (step S9).

Further, returning to step S4, C _e1 > C _e2
If it is, it is determined whether or not the cell is C _e1 (step S1).
0), in the case of the cell C _e1, with priority cell C _e1 connected to the desired outgoing line link (step S11). If a cell C _e2 stores the deflection historical data to deflection cell C _e2 (step S12). With the above configuration, when the switching algorithm is used, the cell with the highest priority with respect to the delay will be the one with the desired priority as long as it does not conflict with the cells with the same priority in the nodes on the way. It will reach the destination node via the route and be sent out of the switch.

When the above-mentioned Rule 1 is used in determining the routing direction, it is the shortest path, and even when Rule 2 is used, it is the shortest path. In the case of the optical switch, the processing time at the node is required to be short, so that the rule 2 with a small calculation amount may be effective. High-priority cells are less likely to be deflected and reach their destinations even with such heavy traffic, while lower-priority cells are more likely to be deflected to their destinations. It may disappear. Even if some delay such as data call is allowed, it is meaningless if the destination cannot be reached. In addition, the low priority cells drifting inside the switch prevent transmission of a newly arrived cell from the outside into the switch, which reduces the throughput of the switch. Furthermore, since the waiting time until the cell is transmitted into the switch increases, the delay time of the high priority cell also increases, and the effect of the priority-based control may be diminished.

Therefore, each cell is provided with a counter for recording the number of deflected links. When the value of the counter exceeds a certain threshold, the priority is raised. With this configuration, it is possible to prevent the generation of a cell that cannot reach the destination and cannot leave the switch. Furthermore, in the above-mentioned basic configuration, the number of frequency channels used is four. However, in practice it may be possible to use some more channels. Therefore,
As shown in FIG. 5, it is possible to construct a logical hierarchical structure by providing a filter that allows a certain frequency to pass through the node and creating a link that jumps over the node.

Hierarchization in this way has the following advantages. (1) When there is a distance from the transmission node to the destination node, the number of hops is reduced by performing routing in the upper layer, and the delay time can be shortened. (2) When a node in the hierarchy has a failure, if it passes through the node in the upper layer, the failure location can be avoided without making a detour.

Next, layering will be described. For example,
The node configuration is as shown in FIG. That is, FIG.
In the node configuration in, the filter 2 on the input lines 1 and 2
By adding the optical couplers 1 and 22 and the optical couplers 23 and 24, and the optical couplers 25 and 26 to the output lines 8 and 9, layering can be achieved.

If the function of the filter / coupler is changed, it is naturally possible to use three or more multiplex signals. The number of switch through stages can be increased by increasing the multiplicity. The present invention is not limited to the above embodiment,
Various modifications are possible based on the spirit of the present invention, and they are not excluded from the scope of the present invention.

[0047]

As described in detail above, according to the present invention, the following effects can be achieved. According to the present invention, the cell having the highest priority with respect to the delay arrives at the destination node through the route as desired, unless the node having the same priority conflicts with the cell having the same priority in the outgoing link. The switch can be sent to the outside of the switch, and high-speed switch processing can be performed with a simple configuration.

If rule 1 described above is used when determining the routing direction, it is the shortest route, and even if rule 2 is used, it is the shortest route. In the case of the optical switch, the processing time at the node is required to be short, so that the rule 2 with a small calculation amount is effective. In addition, each cell has a counter for recording the number of links hopped. If the value of the counter exceeds a certain threshold, the priority is raised.

With this configuration, it is possible to prevent the occurrence of cells that cannot reach the destination and cannot leave the switch.

[Brief description of drawings]

FIG. 1 is a diagram showing a basic configuration of a multimedia optical switch showing an embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of a node of an optical switch.

FIG. 3 is a diagram showing an internal configuration of a node when performing O / E conversion and switching in the node of FIG.

FIG. 4 is a flowchart of data according to the route selection control method of the optical switch of the present invention.

FIG. 5 is a diagram showing an example of layering the optical switch of the present invention.

FIG. 6 is a diagram showing an internal configuration of a node in the layering of the optical switch of the present invention.

FIG. 7 is a diagram showing a structure of a conventional 6 × 6 MSN.

FIG. 8 is a diagram showing a state of a relative address of a conventional 6 × 6 MSN.

FIG. 9 is a diagram showing a conventional routing direction (rule 1).

FIG. 10 is a diagram showing a conventional routing direction (rule 2).

[Explanation of symbols]

1 input line (row) 2 input lines (columns) 3,4 Optical / electrical converter (O / E) 5 Routing switch 6,7 Electric / optical converter (E / O) 8 output lines (lines) 9 output lines (columns) 10, 11 Connection line to the outside of the network 21,22 filter 23, 24, 25, 26 Optical coupler

─────────────────────────────────────────────────── ─── Continuation of front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) H04Q 3/52 G02F 1/313 H04B 10/02

Claims (4)

(57) [Claims] 1. Two incoming links and two outgoing links
And the direction of the link alternates between rows and columns
Yes, and its link has an annular structure, slot period
A network system including a system,
The node with the incoming links and two outgoing lines links of this as a basic unit, the path selection method of controlling an optical switch configured the nodes in a grid of m × m, is transmitted to the (a) incoming link, Fixed to a certain size
In order to select the outgoing link in advance for cells that are packets , each cell has a priority level that represents its delay requirement.
Then, based on this, the output line
When there is a conflict of
Send high priority cells to desired outgoing links and set priority
Low priority cells are sent to another link and given a priority to be deflected ; and (b) when the outgoing links conflict with the cells transmitted to the incoming links, output to the outgoing links according to the priority. An optical switch path selection control method comprising: 2. The route selection control method for an optical switch according to claim 1, wherein when the outgoing links conflict with each other, the cell having a lower priority is deflected and the deflection data is recorded in the cell. Optical switch path selection control method. 3. The optical switch path selection control method according to claim 1, wherein when the priorities are the same, cells are randomly output to an outgoing line link and not output to a desired outgoing line link. A route selection control method for an optical switch, wherein the deflection data is recorded in a cell. 4. The optical switch path selection control method according to claim 2, wherein the deflection data is counted,
An optical switch path selection control method, wherein the priority is raised when the counter value exceeds a certain threshold value.