JPH11163867A - Band control system for local area network - Google Patents

Abstract

PROBLEM TO BE SOLVED: To ensure a band of each communication in the case of using part of links in common in a switch LAN using a fiber channel and conducting one communication of pluralities of classes. SOLUTION: A band management server 5 that manages a band of communication of pluralities of FC nodes 2, 3 is provided in a fabric 1 and band management client sections 10, 11 are provided to each of the nodes 2, 3. Each of the nodes 2, 3 makes a band request including a communication destination and a request band to the band management server 5 via band management client sections 10, 11 prior to the communication. Upon the receipt of the band request from the band management client sections 10, 11, the band management server 5 detects presence of contention of the communication by each node on a link 16 depending on the communication destination. Then the opening/ deletion of connection of a link is instructed to the band management client sections 10, 11 based on the detection result and the value of request band M included in the band request.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a communication band control method in a local area network (LAN).

[0002]

2. Description of the Related Art ANSI (American Nati)
Online Standard Institute (American Standards Institute) and the Fiber Channel Class 1 service standardized by the ISO are bandwidth occupied services, so that stable data transfer can be performed at a high transfer rate. No link on the path can be used for other communications.

[0003] In other words, when a plurality of class 1 communications sharing a part of the links are to be performed simultaneously (typically, communication with the outside via a router or cascade connection of switch fabrics) is performed. The standard does not provide any means for fairly solving class 1 connection conflicts. Therefore, even if intermittent communication is established, once a connection is established, an available communication band cannot be allocated to another connection unless the node requesting the connection deletes the connection.

The connection request receiving mechanism is stochastic and is accepted if the link is vacant, but if the link is already in use, it is rejected and the request is repeated only until it is accepted. Therefore, the time until the connection is accepted is undefined, and it is difficult to satisfy the communication quality such as the maximum delay amount, the delay variation amount, and the minimum bandwidth guarantee necessary for a communication application represented by multimedia communication. For this reason, the first measure is to comply with the Fiber Channel standard and to establish a Class 2 connectionless service.
And class 3 are prepared. As a result, frame multiplexing becomes possible, and improvement in the amount of delay can be expected.

As a second countermeasure, an extended version of the Fiber Channel standard FC-PH-2 defines a connection acceptance mechanism called a camp-on function. This function registers a conflicting connection request in a queue in the switch fabric, and sends a message to one end port of the connection so that the connection is disconnected if communication in the established connection is interrupted. This is a mechanism that controls transmission so that the link band is surrendered to another connection. As a result, the time until a connection request is accepted is shortened, and improvement in communication quality at the time of contention can be expected.

[0006]

However, in the case of the first countermeasure, there is a problem that it is difficult to secure a necessary band since aggressive control is not realized in terms of a communication band. Was. Further, in the case of the second measure, if communication is performed without interruption within an established connection, another communication request cannot be accepted, and thus it is difficult to reliably secure a band. was there. Therefore, the present invention provides a fair bandwidth distribution and a bandwidth allocation when a plurality of class 1 communications are simultaneously performed by sharing a part of links in a switched LAN using a fiber channel which is an ANSI / ISO standard. The purpose is to secure.

[0007]

SUMMARY OF THE INVENTION In order to solve such a problem, the present invention provides a bandwidth management server which manages the bandwidth of communication between a plurality of nodes, and a bandwidth management server in which the bandwidth is managed by the bandwidth management server. A bandwidth management client section is provided for each node, and the managed node issues a bandwidth request including a communication destination and a requested bandwidth to the bandwidth management server via the bandwidth management client section prior to communication, and the bandwidth management server receives a bandwidth request from the bandwidth management client section. When a bandwidth request is input, the presence or absence of communication conflict between the managed nodes on the link determined by the communication destination is detected. Based on this detection result and the value of the requested bandwidth included in the bandwidth request, the connection to the link is The deletion is instructed to the band management client unit to perform band distribution control. Further, upon inputting the requested bandwidth from the managed node, the bandwidth management server calculates the maximum duration of the class 1 connection of the managed node, permits the establishment of the class 1 connection, and sends the calculated maximum duration to the bandwidth management client unit. Upon receiving the notification, the bandwidth management client unit instructs the managed node to open a connection, and controls the connection to be deleted within the instructed maximum duration. Further, the bandwidth management server has a detecting means for detecting a link on the connection path which causes the most contention based on the received information upon receiving a bandwidth request from the managed node, and a maximum duration time Tm of the managed node.
ax is a maximum delay time Td previously given as an operation constant of the network, and a required bandwidth Mn from the managed node.
Tmax = Td × Mn / SM using the sum of the required bandwidths SM for the links detected by the detecting means.
Is calculated. Further, the bandwidth management server calculates the maximum duration time Tmax each time a connection establishment permission is given to each managed node, and notifies each managed node of the calculated value. In addition, a switch fabric for exchanging and controlling communication of each managed node is provided in the LAN, and the bandwidth management server is connected to the switch fabric. In addition, the bandwidth management server
When the event information indicating the detection of the deletion of the connection is fetched from the port of the switch fabric, a permission to open a connection is given to other competing managed nodes in a round-robin manner. If the total value SM of the required bandwidth of the link exceeds the bandwidth capacity of the link due to the occurrence of the new bandwidth request, the bandwidth management server rejects the new bandwidth request.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a system to which the present invention is applied. This system is a LAN formed in a star topology with a switch fabric (hereinafter, fabric) 1 as a center.
(Fiber Channel) The node 2 and the FC node 3 are communication terminating units. The FC node 4 has a function of connecting to an external network as a gateway.
Here, in the present embodiment, the bandwidth management server 5 is installed in the fabric 1 for improving the processing efficiency.

The bandwidth management server 5 is interconnected with the F ports 6 to 8 in the fabric 1 through the crossbar switch 9, and each of the FC nodes 2 through the F ports 6 to 8.
Communicates with. A bandwidth management client unit 10 and a bandwidth management client unit 11 are installed in the FC nodes 2 and 3, respectively. Each of the bandwidth management client units 10 and 11 is connected to an N port 12 provided in the FC nodes 2 and 3. , 13 to communicate with the bandwidth management server 5. Note that the entities of the bandwidth management server 5 and the bandwidth management client units 10 and 11 do not always have independent hardware, but can be realized only by a set of tasks defined on software.

[0010] FC node 2 and FC node 3
When trying to start communication with the node 4 by the class 1 service, the FC link 14 and the FC link 15
Although the entire band can be passed, the FC link 16 causes contention between two communications. Therefore, prior to communication, the bandwidth management client units 10 and 11 communicate with the bandwidth management server 5 by the communication application units 17 and 1 of each FC node 2 and 3.
A request for a band suitable for the communication of No. 8 is made by a command.

FIG. 2A shows an example of the bandwidth request command A. The bandwidth request command A includes at least a source port identifier, a destination port identifier indicating the bandwidth management server 5, a reserved destination port identifier indicating a destination to communicate with, and an average bandwidth value M requested by the communication application unit. Can be The bandwidth management server 5 in the fabric 1 that has received the bandwidth request command A determines the FC link constituting the route from the source and destination node ID (reserved destination port identifier), and sends the FC link from another FC node to the FC link. Check if a bandwidth request is coming. In this way, the FC link with the highest contention, that is, the critical link is detected. In this example, FC link 16
Is detected as a critical link.

The bandwidth management server 5 calculates a connection maximum duration (hereinafter, maximum duration) Tmax of a class 1 connection allocated to each FC node in consideration of a required bandwidth from each node for a critical link. Details of this calculation algorithm will be described later. After calculating the maximum duration Tmax, the bandwidth management server 5 transmits a connection opening permission command B to each of the FC nodes 2 and 3 requesting the bandwidth. FIG. 2B shows an example of the connection opening permission command B. The connection establishment permission command B includes a maximum duration time Tm.
ax.

FIG. 3 is a sequence diagram showing an example of a procedure for opening and deleting a connection using commands between the bandwidth management server 5 and the bandwidth management client unit 10 in the FC notebook # 2 in the system configured as described above. It will be described using FIG. Prior to the start of communication, the bandwidth management client unit 10 recognizes a required bandwidth in the communication application unit 17 based on a notification or instruction from the communication application unit 17, and sends a bandwidth request command A including the requested bandwidth to the bandwidth management server 5. Is transmitted (step S1). Upon receiving the bandwidth request command A, the bandwidth management server 5 inputs the required bandwidth, calculates the maximum duration Tmax, and transmits the connection opening permission command B to the bandwidth management client unit 10 (step S2).

Upon receiving the connection opening permission command B, the bandwidth management client unit 10 immediately instructs the N port 12 to start a connection opening procedure, and instructs connection deletion within the maximum duration Tmax specified by the connection opening permission command B. In order to perform this, a timer (not shown) is started and monitoring of the timer is started.

The procedure for establishing a connection between the N port 12 and the N port 21 is completely compliant with the Fiber Channel standard, is simple without bandwidth negotiation, and is a class 1 having a delimiter called SOFc1.
A start data frame is transmitted from the N port 12 (step S3), and an ACK frame is returned from the N port 21 (step S4), whereby a class 1 connection between the FC nodes 2 and 4 is established. . Here, the unit of data transfer of the fiber channel is a series of data frames in the same direction called a sequence (steps S5 to S7). When the timer reaches the maximum duration Tmax, the bandwidth management client unit 10
After the completion of the sequence during the transfer, the N port 12 is instructed to start the connection deletion procedure.

This connection deletion procedure is also completely compliant with the Fiber Channel standard, and the N port 1
2 displays the end of the connection in the header of the last data frame of the sequence and transmits it to the N port 21 (step S8). N port 21 is the AC for the entire sequence.
A frame in which a delimiter called EOFdt is added to the K frame is returned (step S9). This completes the deletion of the connection. Note that the maximum duration time Tm
Communication application unit 1 continues even if ax has not elapsed
If the transmission data is not prepared in step 7, the bandwidth management client unit 10 may delete the connection.

Once the connection is disconnected, the next connection establishment request from the N port 12 is made when the bandwidth management client unit 10 receives the connection establishment permission command B from the bandwidth management server 5 again. In that case, the newly notified maximum duration time T
max ′ is used (step S10). Thereafter, the bandwidth management client unit 10 similarly instructs the N port 12 to start a connection opening procedure, and starts the timer to instruct connection deletion within the maximum duration time Tmax ′ indicated by the connection opening permission command B. Then, monitoring of the timer is started.

In this case, the N port 12 is also a class 1
The start data frame is transmitted to the N port 21 (step S11), and the N port 21 returns an ACK frame (step S12). As a result, a class 1 connection between the FC nodes 2 and 4 is established.

Thereafter, a series of data frames is transmitted from the N port 12 to the N port 21 (steps S13 and S1).
4) After being done, the timer is set to the maximum duration (Tmax ')
Is reached, the bandwidth management client unit 10 instructs the N port 12 to start the connection deletion procedure after waiting for the completion of the sequence during the transfer. Thereby, the N port 12 displays the end of the connection in the header of the last data frame of the sequence and transmits it to the N port 21 (step S1).
5). Then, the N port 21 returns an ACK frame indicating the disconnection of the connection (step S16). This completes the deletion of the connection. Here, when ending the communication, the bandwidth management client unit 10 issues a bandwidth request M to the bandwidth management
Is notified of "M = 0" indicating the value of "0", and the band is abandoned (step S17).

Incidentally, the calculation of the maximum duration time Tmax is performed as follows. That is, a certain FC node n
The maximum duration Tmax is the maximum delay time Td given in advance as an operation constant of the LAN system, the required bandwidth Mn from the bandwidth managed FC node n, and the critical length on the connection path. Total value SM of required bandwidth on link
Is calculated using the following equation. That is,

[0021]

(Equation 1)

FIGS. 4 and 5 show specific numerical values as to how the duration of the connection is allocated as a result of such calculation. 4 and 5 show an example in which three communications compete with one critical link, and the FC nodes 26 and 2 that are the main subjects of the communications are shown.
The required bandwidth from 7, 28 is 200M (mega) b
This is an example in which ps, 100 Mbps, and 200 Mbps are set. In this example, the link bandwidth capacity is 1000 Mbp.
s, and the maximum delay time Td is 30 ms.

In FIG. 4, when the FC node 28 shown in FIG. 4C does not make a bandwidth request, the F node shown in FIG.
A maximum duration Tmax is allocated at the ratio of the required bandwidth between the C node 26 and the FC node 27 shown in FIG.
The Tmax of No. 6 is 20 ms, and the Tmax of the FC node 27 is 10 ms. Under these circumstances, FIG.
When the FC node 28 shown in (c) requests a band to start communication at the time indicated by the symbol in the figure, the band management server 5
Reduces the maximum duration Tmax for the FC node 26 and the FC node 27 to 12 ms and 6 ms, respectively, and sets the maximum duration Tmax of the FC node 28 to 12 ms.

In order to make the maximum duration Tmax follow the competition condition of the communication at each point in time, the bandwidth management server 5 instructs each FC node a value obtained by recalculating the Tmax for each connection opening permission command. It becomes feasible. Thereby, the overall fairness can be always maintained. In addition, in the steady state, the delay caused by waiting for a connection to be established can be suppressed to the maximum delay time Td or less.
It is clear from That is, in order to implement dynamic bandwidth control, the bandwidth management server 5 notifies each FC node of a value obtained by recalculating the maximum duration Tmax every time a connection establishment permission is given to each FC node, thereby performing arbitration. Addition and deletion of participating connections can be performed at any time, and optimal bandwidth allocation is performed.

Next, in the case where the number of bandwidth requesting nodes decreases, the delay may increase momentarily when the state changes, but the maximum duration Tmax distributed to the remaining bandwidth requesting nodes is shown in FIG. The maximum delay time Td can be satisfied by controlling so as to maintain the value just before one round.

That is, in FIG. 5, each FC node 26,
27 and 28 set the maximum duration Tmax to 12 ms and 6 respectively.
(c) during communication while holding at ms and 12 ms.
When the FC node 28 releases the bandwidth at the time indicated by the symbol in the figure, the bandwidth management server 5
The value 12 which is one cycle before the maximum duration Tmax of 6, 27
ms and 6 ms, and the maximum delay time Td during this time is 1
After 8 ms, the maximum delay time Td is restored to 30 ms from the next cycle, and control is performed so that the maximum duration times Tmax of the FC nodes 26 and 27 are 20 ms and 10 ms, respectively.

If the total required bandwidth for the critical link exceeds the bandwidth capacity of the link of 1000 Mbps, the bandwidth distribution based on the calculation of the maximum duration Tmax will be smaller than the requested bandwidth. In this case, since some communication applications do not operate unless a certain band is guaranteed, an operation mode of a completely guaranteed band is required. For this reason, when the total value SM of the requested bandwidths exceeds the link bandwidth capacity on the critical link on the connection path due to the occurrence of the new bandwidth request, the bandwidth management server 5 rejects the new bandwidth request. When the FC node releases the bandwidth and the bandwidth management server 5 captures the event of the detection of the deletion of the connection from the port of the fabric 1 and there are a plurality of other FC nodes that have been rejected, the bandwidth management server 5 It is also possible to give permission to establish a connection to other competing FC nodes in a round-robin manner. As a result, the unused band can be more effectively used.

In this embodiment, the fabric 1
Although the bandwidth management server 5 is provided in any of the FC nodes, it may be provided in any of the FC nodes. As described above, a facility that has a bandwidth management client function in each FC node while conforming to the Fiber Channel (FC) standard.
By adding a facility having a bandwidth management server function to either the node or the switch fabric 1, an unprecedented function of fairly allocating the link bandwidth can be realized even in a class 1 service, and therefore, the stream type communication is frequently used. The application of the Fiber Channel to the multimedia communication to be performed can be enabled.

[0029]

As described above, according to the present invention, since a connection can be established and released at high speed in the Fiber Channel, after transmitting one or a plurality of sequences in units of a sequence composed of a series of frame groups, By disconnecting the connection and giving up the link band to another communication connection, it becomes possible to distribute a plurality of communication bands to one link. That is, a bandwidth management server that manages the bandwidth of communication between a plurality of nodes is provided, and a bandwidth management client unit is provided for each managed node whose bandwidth is managed by the bandwidth management server. A bandwidth request including a communication destination and a requested bandwidth is made to the bandwidth management server through the section, and the bandwidth management server receives a bandwidth request from the bandwidth management client section and receives a bandwidth request from each managed node on the link determined by the communication destination. Since the presence / absence of communication contention is detected, and the bandwidth management client unit is instructed to open / delete a connection to the link based on the detection result and the value of the requested bandwidth included in the bandwidth request.
When a plurality of class 1 communications are performed simultaneously by sharing one link, fair bandwidth distribution of each communication is performed, and as a result, there is an effect that a necessary bandwidth can be secured at the time of communication of the managed node. When the bandwidth management server inputs the required bandwidth from the managed node, the bandwidth management server class 1 of the managed node.
The maximum duration of the connection is calculated, the permission to open the class 1 connection and the calculated maximum duration are notified to the bandwidth management client unit, and upon receiving the notification, the bandwidth management client unit instructs the managed node to open the connection. At the same time, since the connection is deleted within the instructed maximum duration, there is an effect that a bandwidth suitable for the required bandwidth of the managed node can be secured. Further, when receiving the bandwidth request from the managed node, the bandwidth management server detects a link on the connection path on which the most conflict occurs based on the received information, and determines the maximum duration time Tmax of the managed node in advance by the operation constant of the network. Using the maximum delay time Td given by the following equation, the required bandwidth Mn from the managed node, and the total value SM of the required bandwidths for the links detected by the detecting means, Tmax = T
Since it was calculated by the calculation of d × Mn / SM,
Within the maximum delay interval Td, the connection duration of Tmax is repeatedly given. Therefore, the delay can be suppressed to the maximum delay time Td or less, and the link bandwidth can be monopolized within this connection time. And fair distribution of bandwidth according to Further, the bandwidth management server calculates the maximum duration time Tmax each time it gives a connection establishment permission to each managed node, and notifies the calculated value to the managed node, so that dynamic bandwidth control is realized. Therefore, it is possible to add and delete a connection to a link at any time, so that an optimal bandwidth allocation can be performed. Further, since a switch fabric for exchanging and controlling the communication of each managed node is provided in the LAN and the bandwidth management server is connected to the switch fabric, the connection state of the entire network can be obtained easily and quickly. The accuracy and speed of band control can be improved. In addition, when the bandwidth management server captures event information indicating the detection of connection deletion from a port of the switch fabric, it gives a permission to establish a connection to other competing managed nodes in a round-robin manner, so that the unused bandwidth is used. Can be used more effectively. In addition, the total value SM of the requested bandwidth of the link is generated by the occurrence of a new bandwidth request.
When the bandwidth exceeds the bandwidth capacity of the link, the bandwidth management server rejects a new bandwidth request, so that a managed node that does not operate unless a certain bandwidth is guaranteed can be reliably operated.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a system to which the present invention is applied.

FIG. 2 is a diagram showing a format of a bandwidth request command used in the system (FIG. 2A) and a diagram showing a format of a connection opening permission command (FIG. 2);
(B)).

FIG. 3 is a sequence diagram showing an example of a procedure for establishing / deleting a connection in the system.

FIG. 4 is a diagram illustrating an example of distribution of connection continuation time when the required bandwidth increases in the above system.

FIG. 5 is a diagram showing an example of distribution of connection continuation times when the required bandwidth is reduced in the above system.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Fabric, 2-4 ... FC node, 5 ... Band management server, 6-8 ... F port, 9 ... Crossbar switch, 10,11 ... Band management client part, 12,1
3, 21 ... N port, 15-17 ... FC link, 17,
18: communication application unit, A: bandwidth request command, B: connection opening permission command.

Claims (7)

[Claims] 1. A local area network comprising a plurality of nodes having an ANSI / ISO standard fiber channel interface, wherein a bandwidth management server for managing a bandwidth of communication between the plurality of nodes is provided, and the bandwidth is managed by the bandwidth management server. A bandwidth management client unit is provided for each managed node to be managed, and the managed node makes a bandwidth request including a communication destination and a requested bandwidth to a bandwidth management server via the bandwidth management client unit prior to communication, and When the management server receives a bandwidth request from the bandwidth management client unit, the management server detects the presence or absence of contention for communication between the managed nodes on the link determined by the communication destination, and the detection result and the requested bandwidth included in the bandwidth request are detected. Opening a connection to the link based on the value of Dividing instructs the bandwidth management client unit a bandwidth control method of the local area network and performs distribution control of said band. 2. The bandwidth management server according to claim 1, wherein when the requested bandwidth is input from the managed node, the bandwidth management server calculates a maximum duration Tmax of the class 1 connection of the managed node to permit the establishment of the class 1 connection. The calculated maximum duration time Tmax is notified to the bandwidth management client unit, and the bandwidth management client unit receiving the notification instructs the managed node to open a connection and deletes the connection within the instructed maximum duration time Tmax. A bandwidth control method for a local area network characterized by controlling. 3. The bandwidth management server according to claim 2, wherein the bandwidth management server includes a detection unit configured to detect, when receiving a bandwidth request from the managed node, a link on which a conflict occurs most on the connection path based on the received information. The maximum duration Tma of the managed node
x is a maximum delay time Td previously given as an operation constant of the network, and a required bandwidth M from the managed node.
n and the total value SM of the required bandwidth for the link detected by the detection means, Tmax = Td ×
A bandwidth control method for a local area network, which is calculated by calculating Mn / SM. 4. The bandwidth management server according to claim 3, wherein the bandwidth management server sets the maximum duration time Tmax each time it gives a connection establishment permission to each managed node.
And a calculated value is notified to each managed node. 5. The local area network according to claim 1, wherein a switch fabric for switching and controlling communication of each managed node is provided in the local area network, and a bandwidth management server is connected to the switch fabric. Bandwidth control method. 6. The bandwidth management server according to claim 5, wherein when the bandwidth management server fetches event information indicating that a connection has been deleted from a port of the switch fabric, the bandwidth management server permits another competing managed node to open a connection in a round-robin manner. Bandwidth control method for a local area network. 7. The bandwidth management server according to claim 3, wherein when a total bandwidth requirement SM of the link exceeds a bandwidth capacity of the link due to generation of a new bandwidth request, the bandwidth management server rejects the new bandwidth request. A bandwidth control method for a local area network.