KR102235100B1 - System capable of providing a plurality of servers with a backup communication path through a shared bus - Google Patents

Abstract

According to the present invention, a server system includes: communication equipment including connection ports comprising a plurality of down-ports and at least one up-port, and performing a switching function for outputting a data frame to one of the connection ports in accordance with a destination address of the corresponding data frame, with respect to data frames inputted from the connection ports; servers configured to be directly connected to the plurality of down-ports respectively through a communication line to transmit and receive the data frame through the communication line; first type adaptors provided in the servers respectively; and a second type adaptor configured to be connected to the communication equipment. In addition, the first type adaptors and the second type adaptor are connected to backup buses respectively, and the second type adaptor receives a data frame transmitted by a random one of the servers to the backup buses through the first type adaptor provided in itself to provide the same to the communication equipment, thereby leading the data frame to be outputted to one of the connection ports by the switching function. Therefore, the present invention is capable of considerably reducing costs for building a data center.

Description

[System capable of providing a plurality of servers with a backup communication path through a shared bus}

The present invention, in addition to the main communication path for which data services are currently provided, is provided for server computers that are being built for processing information requests or requested services from numerous client devices. It relates to a system capable of providing a spare communication path that can be used in an emergency such as a failure of a communication path connection failure or a connected switching device.

Governments, large corporations, and small public institutions and small and medium-sized enterprises (SMEs) have a large number of server computers (this specification In, it is abbreviated as'server'.) It operates a data center that is connected to the network line by installing it in a concentrated manner.

For these data centers, operating environments such as temperature and humidity are stably maintained so that various data services can be provided without interruption 24 hours a day, 365 days a year. Eliminate the possibility of occurrence of failure.

In addition, for the stability of data service provision, switching equipment having a communication line connected to each server or a connection port to which the communication line is directly connected (this is referred to as a'down port') is provided. (Usually, it is called a'switch'.) Even when a failure (for example, a connection failure of a communication line or an error during the switching operation of a switch) occurs, the service currently provided by the server will continue as it is. Each server is built to have a redundant communication path to the network.

1, in a data center, partially exemplifies servers and switching equipment constructed such that communication paths are redundant, and each of the servers 10 _i , i = 1, 2, 3, ... The main communication line is connected to the down port of the main switch (3 _i , i = 1,2,..), and the reserved communication line is connected to the down port of the designated spare switch (4 _i , i = 1,2,..). Is connected. For the main switch (3 _i ) and the spare switch (4 _i ) to which the servers are connected, generally, a transmission speed of 1 Gbps is provided for the down port to which the server and the communication line are connected, and the line (1) from the network side A switch that provides a transmission speed of 10 Gbps or higher is used for the connected connection port (this is called an'up port').

In FIG. 1, for simplification of the drawing _{, a configuration in which four down ports are provided to each switch (3 i} , 4 _i ) and the same number of servers are connected is illustrated, but in general, the switch is more than the illustrated one. It has a large number of down ports.

In addition, another high-speed switching device, routing device, or front end processor may be provided at the front end of the main switch 3 _i and the spare switch 4 _{i , and each switch 3 i} , 4 _i ) is connected to these different communication equipment through the main line (1) and the spare line (2), respectively, in case of an emergency such as a failure of the main line (1) or the communication equipment connected to the main line (1). It is redundant so that a spare line 2 can be used.

However, as illustrated in FIG. 1, even when a communication line or switch fails, a spare switch (12 _i , i = 1, 2,...) so that the data service currently provided by an arbitrary server can be maintained as it is. Duplicating the communication line for each server with) has the advantage of being able to sustain service through another communication channel by coping with the failure occurring on the current communication channel. There is a disadvantage of increasing the cost for building or maintaining a data center, because a spare switch, which is provided and costs several million won, needs to be preliminarily built in response to each of the main switches.

Moreover, today, the performance of communication equipment and the like is very stable, so that the frequency of occurrence of a failure or the like is extremely rare, so in order to prepare for a case that can occur only one, each _{main switch 3 i as illustrated in FIG. 1 Building the spare switch 4 i} for preliminary use has a problem that the cost-effectiveness is too small.

Accordingly, there is a need for an efficient method capable of providing a spare communication path that can be used in case of a failure of a communication line connection or a failure of a communication device such as a switch at a lower cost.

An object of the present invention is to provide a system capable of providing a spare communication path for many servers while minimizing communication equipment that must be provided for redundancy.

Another object of the present invention is to provide a system capable of further improving the service stability of servers by providing two or more spare communication paths for a large number of servers.

Another object of the present invention is to distribute the load to the spare communication paths for a large number of servers, so that even when a number of failures occur at the same time, the speed of the service provided by the servers through the spare communication path is at an appropriate level. It is to provide a system that allows it to be maintained.

The object of the present invention is not limited to the above explicitly stated object, and naturally includes in the object to achieve the effects that can be derived from the specific and exemplary description of the present invention.

According to an aspect of the present invention, a server system for processing a request for a data service from client devices includes connection ports comprising a plurality of first ports and one or more second ports to which a line from a network side is connected. And a communication device configured to perform a switching function for outputting data frames input from the connection ports to any one connection port according to the destination address of the data frame, and at least one of the plurality of first ports In some cases, a first group of servers individually configured to be individually connected to each other through a communication line to transmit and receive data frames through the communication line, and first type adapters provided in each of the servers of the first group, And a first adapter of the second type configured to be connected to the communication equipment, for example composed of a separate device connected integrally with the communication equipment or connected by a communication line with the communication equipment. In addition, the adapters of the first type and the first adapter are respectively connected to a spare bus, and the first adapter is By receiving a data frame transmitted to the spare bus through an adapter and providing it to the communication device, it is output to one of the connection ports by the switching function.

In an embodiment according to the present invention, the server system includes N _SW (N _SW is 1 or more) communication devices _{configured to perform the same operation by being configured identically to the communication devices, and the N SW} communication devices. It is connected on the attached N _SW-server group, N _SW of telecommunications equipment to the at least one, and a second adapter of said second type is configured to send and receive data frames may be further comprising: via the spare bus. In this embodiment, servers belonging to each group of the _{N SW} server groups are directly connected to at least some of the plurality of first ports provided in any one of the _{N SW communication devices through a communication line, and the connected} It is individually configured to transmit and receive data frames through communication lines. And, the first adapter and the second adapter, for data frames received from the spare bus, a data frame having a source address that matches any one of addresses assigned to relay by itself is connected to It is provided to the existing communication equipment, and data frames that are not are discarded.

In an embodiment according to the present invention, each of the first adapter and the second adapter, a communication device to which it is connected, does not normally perform the switching function, or communicates with a communication device to which it is connected. When the line connection is bad, it notifies another adapter of the second type through the spare bus of its relay failure.

In an embodiment according to the present invention, the second type of adapters connected to the spare bus, including the first adapter and the second adapter, respectively, in a data cycle repeatedly formed on the spare bus. , An address allocated to identify any adapter of the second type that cannot transmit a specific signal in a predetermined specific section, and to take charge of relaying all source addresses of data frames that can be received through the spare bus. It is possible to designate a relay disabled for any adapter identified above on the allocation map. In addition, in this embodiment, if the identified arbitrary adapter is an adapter that has formed the data cycle on the spare bus, the specific signal is transmitted to the next section of the specific section among the second type of adapters. A third adapter of the second type designated to form a data cycle on the spare bus.

In one embodiment according to the present invention, the address allocation map is set for each of the adapters of the second type, and in the address allocation map, the adapters of the second type are relayed to all the source addresses. The order in charge is specified and the information can be recorded. In this case, the first adapter and the second adapter each have the earliest order of designation in which the relay assignment map allocates itself as a relay for the source address of an arbitrary data frame received from the spare bus. Even if it is not, when the second type of random adapter assigned as a relay in charge of relaying in an order preceding the designated order is designated in the relay assignment map because relaying is not possible, the communication device to which it is connected to the arbitrary data frame By providing it as, the relay is handled by himself/herself.

In an embodiment according to the present invention, the server system is configured to further include a second spare bus. In this embodiment, the adapters of the first type and the first adapter are connected together to the spare bus and the second spare bus, and one of the spare buses is selectively used. In this embodiment, the adapters of the first type and the first adapter respectively check the load levels of the spare bus and the second spare bus to select a spare bus having a lower load level, or, The reserved bus that can be used at the current point of time can be selected and used for data frame transmission.

A system capable of providing a spare communication path to a plurality of servers through a shared bus, according to the above-described present invention or at least one embodiment of the present invention described in detail with the accompanying drawings below, includes a data frame or a packet. It is possible to provide a backup communication path for servers to which a communication line is connected to the device without preliminary provision of a communication device for switching or routing the device. Accordingly, it is possible to greatly reduce the cost required in constructing or maintaining a data center with many servers, network lines, and necessary communication equipment.

In addition, the present invention is much more improved than the prior art having a switch as a certificate for a spare communication path that is instead used when the main communication path used by servers is unavailable due to a failure of communication equipment such as a switch. It can provide stability. Considering the situation in which the frequency of failures such as switches is extremely low due to the manufacture of communication equipment, the evolution of production technology, or the improvement of the quality of parts used, the down ports of a large number of switches are viewed on the shared bus that provides a spare communication path. When connected through the device according to the invention, communication equipment is multiplexed against failure by the number of connected switches. This can significantly reduce the possibility that the use of a spare communication path becomes impossible in the event of simultaneous failures of switches. In addition, according to an embodiment of the present invention, when a bus used as a spare communication path is duplicated, stability of various services provided through servers can be maximized. Of course, the present invention can also achieve such maximization of stability without having a preliminary switch.

1 illustrates a general system construction method for redundant communication paths for the stability of service provision for servers constituting a data center,
2 is a diagram illustrating a part of a server system constructed by connecting servers and switches to each other according to an embodiment of the present invention,
3 is a block diagram of the configuration of an adapter for providing a shared bus as a spare communication path by being respectively connected to a server and a switch according to an embodiment of the present invention,
4 is an exemplary diagram illustrating that, according to an embodiment of the present invention, when a time division multiple access method is applied to a shared bus for providing a spare communication path, a random timeslot is occupied and a data frame is transmitted through the bus. Is shown as,
FIG. 5 is a diagram illustrating a relay assignment map showing an example in which adapters connected to a switch quadruple servers to relay data frames and allocate them according to an embodiment of the present invention,
6 shows an example of a server system constructed by redundantly constructing a spare bus to be used when a failure occurs in a main communication path of servers according to an embodiment of the present invention,
7A and 7B show, in the header of a data cycle, an information field necessary to select and use one bus in consideration of the current load applied to each bus in the redundant spare bus according to an embodiment of the present invention. It is a diagrammatic representation of an example of being used and the process of selecting a bus to use based on the information field,
FIG. 8 is a data cycle in order to enable a subject forming a data cycle on a bus for occupancy arbitration on a bus used as a spare communication path to be seamlessly switched according to an embodiment of the present invention. The beacon zone secured in the header of and the number of cases in which signals are loaded in the beacon zone are exemplified.
FIG. 9 schematically illustrates one method of determining, by an adapter connected to a switch, a data frame relay impossible state of the connected switch, according to an embodiment of the present invention,
10 is a diagram illustrating a part of a server system in which an adapter for relaying through a preliminary communication path of a data frame is partially provided for switches according to another embodiment of the present invention.

Hereinafter, various embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

In the following description of the embodiments according to the present invention and in the accompanying drawings, the same numbers added refer to the same components unless there are special circumstances. Of course, for convenience of description and aid in understanding, the same components may be added with different numbers as needed.

2 is a diagram illustrating a part of a server system that comprehensively provides data services, constructed by connecting servers and switches, which are switching devices, according to an embodiment of the present invention. The server system illustrated in FIG. 2 includes a plurality of servers (20 _i , i = 1, 2, 3, ...) operating independently of each other for processing requests for data services from client devices, and Except for a specific port among the plurality of down ports, each down port includes switches (11 _i , i = 1, 2, ..) connected to each server through a communication line. Each of these switches (11 _i , i = 1, 2,..) is configured for a data frame input to any one of the available connection ports according to the destination address of the data frame. It performs a switching function that outputs to the connection port.

In addition, the server system of FIG. 2 _{is individually mounted or provided on the servers (20 i} , i = 1, 2, 3, ...) to transmit and receive data through the shared spare bus 100. Type of adapters (21 _i , i = 1, 2, 3,...) and _{connected to the specific down port of each switch 11 i} to communicate with the corresponding switch, while at the same time, the spare bus 100 _{It is configured to include adapters of the type (22 i} , i = 1, 2,..) capable of transmitting and receiving data through.

In the present specification, for the type of adapter (21 _{i ) provided in each server (20 i ) and the type of adapter (22 i} ) connected to a specific down port of each switch (11 _i ), to be referred to separately from each other. When necessary, they are referred to as'terminal adapters'and'gateadapters' respectively, but when there is no need to distinguish them, they are simply referred to as'adapters'.

The server system partly illustrated in FIG. 2 is not preliminarily built with switches, which are switching equipment, compared to the conventional system illustrated in FIG. 1. That is, it is used to provide a main communication path for each group of servers to which all the switches 11 _{i are connected.} In a configuration in which servers are connected to routing equipment other than switching equipment, it can be constructed as illustrated in FIG. 2.

Of course, a group of servers connected to an arbitrary switch (or routing device) may be a number of servers less than the number of down ports that can be connected to the switch (a number less than the number of down ports provided in the switch).

3 is a block diagram of the configuration of _{the adapters 21 i} and 22 _i according to an embodiment of the present invention. The adapter (21 _i , 22 _i ) participates in a discovery cycle formed on the spare bus 100 and, when the discovery cycle ends, is allocated to itself according to the discovery result, a data cycle A bus controller 202 that transmits data by occupying a bus in the time period (in this specification, this time period is referred to as a'timeslot'), and the bits of the data transmitted by the bus controller 202 are , Transceiver that converts and transmits a physical signal required for transmission through the shared spare bus 100, converts the signal detected by the spare bus 100 into data bits, and delivers it to the bus controller 202 ( transceiver) 201 and an interface controller 204 capable of sending and receiving data frames with the device according to a data transmission/reception method with a connected device (server or switch), and the bus controller 202 and the interface controller 204 It is configured to include a buffer 203 that provides a temporary storage space for transferring data frames between.

Among the adapters 21 _i and 22 _i , preferably each bus controller 202 of the gate adapters 22 _i performs a cycle of discovering all adapters sharing the spare bus 100 with each other. And, the nodes connected to the bus discovered in the discovery process, i.e., adapters (21 _i , 22 _i ), are consecutive timeslots that allow them to exclusively occupy the bus in order according to the time-division method without collision with each other. It has the ability to repeatedly form data cycles on the bus. In the present specification, when it is necessary to specifically classify and refer to a bus controller having such a function, it is referred to as a'master controller'.

Of course, depending on the embodiment, a bus controller of a terminal adapter other than a gate adapter may be implemented as a master controller. Therefore, hereinafter, the description will be made on the assumption that the bus controller of the gate adapter is the master controller, but the technical principle of the present invention will be described in detail with an example hereinafter for an embodiment in which the bus controller of the terminal adapter is implemented as a master controller. Since the and ideas can be applied as they are, the scope of the rights should be interpreted as the scope of the rights regardless of what type of adapter the master controller is implemented in, unless explicitly excluded from the claims.

In addition, _{arbitration of access to the spare bus 100 shared by the adapters 21 i} and 22 _i in a time-division manner is only one example for explaining the principles and technical ideas of the present invention. Naturally, a method of arbitrating bus occupancy in a manner other than a time division method, for example, a bus arbitration method that allows transmission at any time such as CSMA/CD (Carrier Sense Multiple Access/Collision Detection), or by frequency or code division. A bus arbitration method may be applied to the spare bus 100. Of course, in the case of adopting such a multi-access arbitration scheme, _{the bus controller 202 of the adapters 21 and 22 i} occupies the bus in accordance with the arbitration scheme. In the case of a bus arbitration method not based on time division, a master controller having a function of forming a data cycle on the bus for time division arbitration of bus occupancy is naturally unnecessary.

When a situation in which normal transmission of data is impossible such as a communication equipment failure on the main communication path or a bad connection of a communication line occurs, the gate adapter 22 _i is installed in the spare bus 100 that can be used as a spare communication path. If a plurality of units are connected, there are a plurality of master controllers that perform formation of discovery cycles and data cycles on the spare bus 100. Therefore, when all of them perform their functions, signals such as data cycles collide on the bus, making normal use of the spare bus impossible.

In order to prevent such a case, a value indicating a priority for activation of the function (hereinafter, this value will be referred to as'activation sequence number') is individually set in _{the gate adapter 22 i, and} On a shared bus, only a single master controller can perform its functions at all times. The activation sequence may be set through an element such as a dip switch provided in the corresponding adapter. For example, by setting the gate adapters to different values sequentially starting from 0 by each dip switch, each master controller reads the value of the dip switch and responds to the operation in which it must form a cycle on the bus. You can check the activation sequence for.

In addition, only the master controller that has confirmed that the activation sequence set in the dip switch is the fastest value, for example, 0, forms a discovery cycle and a data cycle on the shared spare bus 100 as described above. In this way, a master controller that is performing a data cycle on the spare bus 100 at the present time is specifically referred to as a'current master controller'. Then, the master controller to which the next activation sequence is assigned is formed on the bus by inheriting the cycle required at the present time when all master controllers in the activation sequence preceding itself fail to perform their functions. This will be described in detail later.

In the server system partly illustrated in FIG. 2, each server 20 _i has a failure _{in the switch 11 i} , in which the communication line is connected to each down port in a one-to-one manner, according to the same method as in the prior art, or When recognizing that there is a communication failure in the main communication path due to a problem such as a bad connection of the communication line, the data frame to be transmitted is transmitted to the terminal adapter 21 _i connected to the user. The data frame transmitted by each server 20 _i generally includes a specific type of address in its header, for example, a destination and a source address in the form of a MAC (Media Access Control) address.

_{The data frame transmitted to the terminal adapter 21 i} connected to (or equipped) by any server that recognizes a communication failure through the main communication path is received by the interface controller 204 in the corresponding terminal adapter and the buffer It is stored in the transmission space of 203. In an embodiment according to the present invention, when the interface controller 204 stores the received data frame in the transmission space of the buffer 203, the source address is assigned to the interface controller, for example It can also be saved by replacing it with a MAC address.

If the data frame is stored in the transmission space of the buffer 203, the bus controller 202 in the adapter has the occupancy of the bus from the signal detected and transmitted by the transceiver 201 from the spare bus 100. If it is possible to check whether it is a possible time point, and if it is an occupiable time point, for example, in the data cycle formed on the bus by the current master controller, as illustrated in FIG. 4, if the time slot assigned to it starts starts (401). , The data frame stored in the transmission space is read and transmitted to the transceiver 201 in the form of a bit string, so that the signal by the data frame is loaded on the spare bus 100.

In this way, the operation of the bus controller 202 to transmit a data frame received by the interface controller 204 and stored in the transmission space of the buffer 203 to the spare bus 100 through the transceiver 201 is performed by the switches 11 _The same is done in the gate adapters 22 _{i respectively connected to i ).}

When a signal by a data frame is loaded in an arbitrary timeslot, the corresponding timeslot becomes longer than the _{original allocated time width (T AlotTW ).} In order to notify the end of the corresponding timeslot, the bus controller 202 transmits a termination identifier 411 consisting of a special bit pattern following the transmitted data frame. The bus controller of each adapter that detects the bit pattern of the termination identifier recognizes that the corresponding timeslot is ended and the next timeslot is in progress.

In this way, the data frame 410 loaded on the spare bus 100 is received by all the adapters 21 _i and 22 _i connected to the bus 100. That is, a series of bit strings obtained from signals detected by the transceiver 201 of each adapter are configured as one data frame by the bus controller 202 of the adapter and stored in the receiving space of the buffer 203. Then, the interface controller 204 in the adapter determines whether to delete the data frame stored in the receiving space or to transmit it to a device (server or switch) connected to it.

In an embodiment according to the present invention, the determination of such a dish is made based on an address, for example a MAC address, carried in a header of a data frame. This will be described in more detail below.

If the interface controller 204 is a component in the terminal adapter 21 _i , the interface controller has the destination address of the data frame to be read from the receiving space of the buffer 203, a data frame transmitted from the server to which it is connected. If it matches the address given as the source address in the server, that is, the address given to the communication module of the server (depending on the embodiment, the interface controller itself, that is, the address assigned to the terminal adapter), the corresponding data frame is read from the receiving space. It sends it to the server connected to itself, and deletes it if it doesn't match.

By the way, a data frame transmitted by an _{arbitrary server (20 i ) to the terminal adapter (21 i} ) connected to it or attached to it in order to use the spare communication path due to the failure of the main communication path used by the server (20 i) Is, since the destination address cannot be for other servers in the server system (and, depending on the embodiment, it may not be the address given to the terminal adapters each attached to the servers), All of the interface controllers 204 in the other terminal adapters delete the data frame from the receiving space of the buffer 203.

On the other hand, if the interface controller 204 is a component in the gate adapter 22 _{i attached and connected to an arbitrary switch 11 i} , it reads from the receiving space of the buffer 203 with reference to the relay allocation map set to itself in advance. It decides whether to discard the transmitted data frame or to forward it to the connected switch. 5 shows an example of a relay allocation map 50 constructed according to an embodiment of the present invention, in which four gate adapters are connected to a shared bus constructed for preliminary use, and the gate adapter Each of the attached switches is equipped with 16 down ports (of which one specific down port is assigned for connection to the gate adapter), and each down port is targeted at a part of the server system connected one-to-one through a communication line with the server. It is exemplified. Of course, if a spare communication path other than the 60 servers connected to these four switches is built through a separate spare bus, the interface controller in the gate adapter attached to each of the switches constituting a part of the server system Also, a relay assignment map configured in the same manner as in FIG. 5 will be set.

The relay assignment map 50 illustrated in FIG. 5 may be set through an I/O interface separately provided in the gate adapter, or alternatively, relay assignment to a data frame of a specific format predetermined with the interface controller of the gate adapter. The map information can be loaded and set by transmitting the address assigned to the gate adapter, for example, a MAC address (more specifically, a MAC address assigned to the interface controller of the gate adapter) or an IP address as a destination.

In an embodiment according to the present invention, the gate adapter 22 _i may be integrally configured with the switch. _{In this case, the interface controller is connected in a manner requested by the internal bus of the switch 11 i} , not in a communication method required by the down port, and communicates with the processing module in charge of switching or communicates with the internal controller. In this embodiment, information on the relay assignment map may be set on a menu screen provided by the internal controller through an I/O interface provided in the switch.

As for the identification number of the gate adapter in the relay assignment map 50 illustrated in FIG. 5, the activation order as described above may be used as it is, or a unique identification number that uniquely identifies each gate adapter is one-to-one with the activation order. Correspondingly, it may be specified separately. A method for _{the interface controller 204 in the gate adapter 22 i} to determine whether it is to relay by referring to the relay assignment map 50 configured as illustrated in FIG. 5 is as follows.

The interface controller 204 finds a source address of a data frame to be read from the reception space of the buffer 203 in an address column of the relay assignment map 50. And, when the corresponding address is found, the identification number indicating the current'relayable state' is determined starting from the first for the gate adapter identification number assigned to the address. And, if the identified identification number is the same as the identification number assigned to itself, the interface controller 204 reads the data frame from the buffer 203 and transmits it to the connected switch. And, if the identified identification number is different from the identification number assigned to itself, the interface controller 204 erases the data frame from the receiving space.

For example, if the source address of the read data frame _{is any one of Addr 31} to Addr ₄₅ , a gate adapter assigned with identification number 3 (e.g., to the same switch as the server that transmitted the data frame to the spare bus) Only the interface controller of the connected gate adapter) transmits the data frame to the switch (501), and the interface controllers of the gate adapter to which identification numbers 0 and 1 are assigned discard the data frame. In the example of FIG. 5, since the interface controller of the gate adapter assigned the identification number 2 cannot relay the data frame for some reason (502), it cannot read the data frame from the receiving space of the current buffer. Even if there is, the data frame is discarded because the address assigned to it is not the source, and if such an operation itself is impossible due to a failure of the corresponding gate adapter, of course, the data frame cannot be relayed to the connected switch. .

When any gate adapter connected to the shared spare bus 100 is in a state in which the data frame received from the spare bus cannot be relayed to the switch, other gate adapters recognize it and, as illustrated in FIG. Likewise, in each relay assignment map 50, the identification number of the corresponding gate adapter is designated as a'relay disabled state' (502). How the other gate adapter recognizes that any one gate adapter is in a'relay disabled state' will be described in detail later.

As illustrated in FIG. 5, when the gate adapter of the identification number 2 is designated as a relay disabled state (502), the source of the data frame to be read from the reception space of the buffer 203 by the interface controller 204 of the other gate adapters If the address is _{any one of Addr 31} to Addr ₃₅ , the interface controller of the gate adapter of identification numbers 0 and 1 discards the corresponding data frame through reference of the relay assignment map as described above, and the gate adapter of identification number 3 The interface controller relays the data frame to the switch in place of the gate adapter of identification number 2 (503).

Therefore, as illustrated in FIG. 5, if four gate adapters are connected to one spare bus, even if a failure occurs in three switches at the same time, the faulty switches can be prevented through the gate adapter connected to the other switch. Data frames from servers to which the main communication path is connected can be relayed to a switch in normal operation. In other words, it is possible to secure quadruple stability for a system built with servers without having any spare switch for use only in case of emergency such as failure, etc., which is provided with a spare switch per main switch. The stability is improved even more compared to the conventional redundancy. Of course, this stability can be further doubled by increasing the number of gate adapters connected to the spare bus 100.

On the other hand, if a large number of servers are connected to one spare bus to secure a spare communication path, the possibility is very low, but when more than one switch to which the main communication path is connected fails at the same time, the amount of data that will use the spare bus is By increasing, more than a reasonable level of delay may occur in the transmission over the spare bus.

Accordingly, in an embodiment according to the present invention, as illustrated in FIG. 6, the spare bus may be duplicated. Of course, it is also possible to multiplex three or more spare buses. In the case of multiplexing the spare buses in this way, each adapter is provided with a transceiver as many times as the multiplexed number. Of course, depending on the embodiment, the bus controller may also be provided with the number of multiplexed buses, respectively. Hereinafter, for simplicity of explanation, it is assumed that the spare bus is duplicated.

The activation order of each gate adapter for the first spare bus 100 and the second spare bus 101 may be designated identically or differently. Accordingly, the current master controllers for the first and second spare buses 100 may be the same controller or different bus controllers depending on the embodiment.

The current master controller for each spare bus, when forming a data cycle on the corresponding bus, includes a bus load information field 71, as illustrated in Fig. 7A, in the header of the data cycle. In addition, a beacon zone 70 is additionally secured in the header of the data cycle proceeding by the current master controller. The description of this beacon zone will be described in detail later. Further, depending on the embodiment, another field 72 required for the header of the data cycle may be provided.

The current master controller keeps track of how many timeslots are used for transmission of data frames in each data cycle it is repeatedly advancing to at least one of the first spare bus 100 and the second spare bus 101. To confirm. For example, check the number of timeslots used for transmission of a data frame in 10 data cycles prior to the current data cycle, and information on the average value of the number of timeslots identified (711), for example , A value representing the percentage of the average number of timeslots used with respect to the number of timeslots per data cycle is recorded in the load information field 71 of the data cycle to be formed in the spare bus at the present time.

By recording the information in the load information field, each adapter can grasp the degree of load on the data cycle formed on the spare bus. In particular, it is possible to know which one of the first spare bus 100 and the second spare bus 101 has a larger current load. The greater the load on the bus, the longer the delay until the transmission of the data frame through the spare bus is completed.

_{Accordingly, the bus controller 202 of each adapter 21 i} and 22 _i connected to both the first spare bus 100 and the second spare bus 101 is, as illustrated in FIG. 7B, each spare bus 100, 101 ), when the data cycle starts, the information in the bus load information field is checked, and which spare bus is in the smaller load state (S73). And, when it becomes its own time slot on the spare bus under a smaller load, if there is a data frame in the transmission space of the buffer 103, the frame is transmitted to the spare bus through the transceiver 201 (731). .

In another embodiment according to the present invention, the load information field may not be included in the data cycle. In this embodiment _{, the bus controller 202 of each adapter 21 i} and 22 _i continuously monitors the number of timeslots used per data cycle for each spare bus, as described above, so that the load of each bus It is also possible to determine the degree and use a spare bus with a low load strength at the present time.

In another embodiment according to the present invention, when there is a data frame to be transmitted on the spare bus, a bus that can be used first among the connected spare buses 100 and 101 may be used. That is, if the time-division bus occupancy arbitration method is applied to each spare bus, it is possible to use a bus in which its own timeslot starts first on the spare bus.

On the other hand, when the spare bus is redundant as shown in FIG. 6 _{, the bus controller 202 of each terminal adapter 21 i} is configured with any live signal ( This'live signal' will be described later. If) is also not detected, the spare bus is determined to be unavailable and another spare bus is used.

On the other hand, since a single or multiple spare buses are intended to provide a spare communication path for a large number of servers, when a state in which transmission of data frames through the spare bus is impossible is made, it is possible that a large number of servers It will be that the redundant communication paths are closed all at once.

Accordingly, in an embodiment in which the time-division bus arbitration method is applied to the spare bus as described above, the function of the current master controller that informs that the data cycle that enables each time slot to be sequentially recognized is started is very important. For this reason, for the current master controller, it is sequentially backed up according to the activation sequence as mentioned above. That is, when a failure occurs in the master controller functioning as the current master controller, the master controller to which the activation sequence indicating the next sequence is assigned functions as the current master controller. This switching of the current master controller is based on a signal placed in the beacon zone of the data cycle.

8 illustrates a structure of a beacon zone secured in a header of a data cycle according to an embodiment of the present invention, wherein the beacon zone 80 includes beacon fields as many as the number of master controllers connected to a corresponding spare bus. (81) and a guard field (82). When the current master controller is switched, the guard field 82 is a field for securing a switching time required for a new master controller, which has become a current master controller, to proceed in place of the currently ongoing data cycle, and is applied. Depending on the embodiment, the beacon zone may not include a guard field.

And, for each beacon field, a master controller to which an activation sequence that matches the order in the beacon zone is assigned a special signal or a signal corresponding to a specific bit pattern (for example, "10" or "1110") (this In the specification, this is referred to as a “live signal.”) is transmitted through a transceiver to be loaded on a bus.

Therefore, when each master controller is in a normal state without any obstacles in transmitting and receiving signals through the spare bus, a live signal is loaded in the beacon field assigned to the corresponding master controller, and each master controller has each beacon field. It is possible to check whether the other master controller is currently faulty through detection of the live signal in.

If the master controller, which has the highest priority in the activation sequence, is functioning as the current master controller and there is an error in signal transmission and reception with the spare bus, so that no signal is loaded in the corresponding beacon field 811, an activation sequence indicating the next sequence is given. After the master controller transmits a live signal to its beacon field, the data cycle is inherited and proceeds _{at the time point (t m1) when the beacon zone ends.}

The master controller that proceeds on the spare bus instead of the data cycle (i.e., becomes the current master controller) starts from the first beacon field and immediately follows the last beacon field in which the live signal is not continuously loaded, and sends the live signal to the beacon field. It is the bus controller that is sent to. That is, in the beacon fields, if a beacon field in which a live signal is not loaded is not continuous from the beginning and appears in the middle (812), the master controller that transmits the live signal to the next beacon field does not proceed with the data cycle instead. . In this way, even if the bus controller, which is the current master controller, fails, the other master controller automatically takes over and allows the data cycle to continue on the spare bus, causing any problem in bus use of the adapters connected to the spare bus. Do not. Therefore, sufficient stability is ensured for the function of the bus as a spare communication path.

On the other hand, as illustrated in FIG. 8, if, in the beacon fields 81, a beacon field in which any master controller did not load a live signal (in the example of FIG. 8, the activation sequence indicating the highest priority order that was functioning as the current master controller If there is a first beacon field 811) in which the assigned master controller does not carry a live signal, other master controllers notify each of the interface controllers in the same gate adapter, so that in the relay assignment map 50 as described above, live For a gate adapter with a built-in master controller that does not load a signal on the bus, more precisely, an identification number assigned to the gate adapter (this may be an activation sequence number as mentioned above, or a separate one-to-one correspondence with the activation sequence number). It may be the number of.) To be designated as'relay disabled'.

Accordingly, for servers designated so that the gate adapter including the master controller that has not transmitted the live signal to the allocated beacon field is in charge of relay, other gate adapters are sequentially designated in the relay allocation map 50 You will be in charge of relaying.

For seamless switching of the current master controller, the beacon fields 81 secured in the header for each master controller do not replace the current master controller, which proceeds the data cycle instead of the bus. In some cases, the relay allocation map 50 is used only for state change of the corresponding gate adapter. That is, as illustrated in FIG. 8, while the master controller of the gate adapter to which the current activation order 0 is assigned is functioning as the current master controller, the master controller of the gate adapter to which the activation order 2 is assigned in the beacon zone 80 is If it fails to transmit a live signal to its beacon field 821, the current master controller forming the data cycle is not replaced, and the interface controllers of other gate adapters are notified of the fact from the master controller in the adapter, and By only designating the activation sequence number 2 as a'relay disabled state' in the managed relay assignment map 50, another gate adapter can take over relaying to the servers allocated to the gate adapter that does not transmit a live signal.

On the other hand, in _{addition to a failure situation in which the bus controller included in the gate adapter 22 i} , that is, the master controller cannot load a live signal in its beacon field, the data frame loaded on the spare bus by the corresponding gate adapter is transferred to the switch 11 _i . There may be other situations where we are unable to relay. For example, an arbitrary gate adapter may _{fail to relay a data frame to the switch due to a failure in the connected switch 11 i} , or a poor connection of the communication line connecting the gate adapter and the switch to each other, etc. The situation may occur. When this happens, the gate adapter cannot normally transfer the data frame to the connected switch. Accordingly, in an embodiment according to the present invention, such a fault condition is detected and the other gate adapter is notified so that it can relay on its behalf.

In an embodiment according to the present invention, the interface controller 204 of the gate adapter, as described above, has addresses 90 that are allocated to it for relay (the identification number of the gate adapter 50 on the relay assignment map 50 is' When a data frame having an address belonging to the address designated as'identification number of the primary relay assignment gate adapter') as the source address is received through the buffer 203, the data frame is relayed to the connected switch and at the same time, as shown in FIG. As illustrated, the address is marked as'in use of a preliminary communication path' (901).

In addition, with this mark, it is checked whether the currently marked address is the last address designated as'in use of a preliminary communication path' in the addresses 90 designated to be relayed by itself. If the current'spare communication path in use' mark is 902, and finally, if the address is marked as'spare communication path in use' (91), the interface controller uses the related servers as the main communication path. It is determined that a failure has occurred in the switch itself to which the communication line is connected, and notifies the bus controller 202 in the same adapter that the relay is impossible (rc2 in FIG. 2 ). In addition, the relay assignment map is marked as'relay disabled' for its identification number.

When there is such a notification, the bus controller 202 does not transmit a live signal to the beacon field allocated to it in the beacon zone 80 as described above in order to notify other gate adapters of its relay failure. Not (p92). Since the live signal is not transmitted to the corresponding beacon field, as described above, each bus controller of the other gate adapters is a value indicating the order of the beacon field in which the live signal is not loaded (ie, a value that identifies the corresponding gate adapter). Is transmitted to the interface controller 204 in the same adapter to inform the'relay failure' (rc2). Then, the other gate adapters update the information of the relay assignment map set to each of them (mark the'relay disabled state' with respect to the gate adapter that has notified the relay failure), and the interface controller 204 in any one of them. ) Refers to the updated relay assignment map, and relays the data frame to be loaded on the spare bus later on behalf of the gate adapter that has not transmitted the live signal.

In the case of a failure such as a connection failure of the communication line connected to the switch, it can be detected through a sensing circuit provided in the interface controller 204 of the gate adapter. For example, the voltage level between both signal lines of the signal lines constituting the communication line, or the voltage level of one specific signal line is not a predetermined level, e.g., the voltage that appears when it is in a floating state. , When detected from the provided sensing circuit, the interface controller 204 determines that relaying to the switch is impossible, and operates in the same manner as when the above-described switch failure is determined. The gate adapter operates normally, but of the connected switch. When relaying of a data frame is impossible due to a failure, as described above, if a live signal is not transmitted to the beacon field, if the bus controller of the gate adapter is the current master controller, the transfer to another master controller proceeds.

However, in this case, since the transfer is unnecessary, in another embodiment according to the present invention, when the current master controller cannot function and when the connected switch is confirmed to be incapable of relaying, it is notified to other master controllers. May be. In this embodiment, for this purpose, a separate relay information notification zone may be allocated to the header of the data cycle in the same manner as the beacon zone. That is, as described above, the master controller of the gate adapter whose connected switch is confirmed to be relay disabled is'relay disabled' in the relay information notification field assigned to itself in the relay information notification zone assigned to the header when the data cycle starts. (P92), and the bus controllers of other gate adapters transmit the information contained in each relay information notification field to a value indicating the order of the notification field (ie, a value that identifies the corresponding gate adapter). Together with, the interface controller 204 is notified. The subsequent process is as described above.

On the other hand, the gate adapter, which notifies other gate adapters through the beacon field or the relay information notification field, of the relay failure of the connected switch, does not relay data frames to the switch, but all data frames loaded on the spare bus. As is, it is received as it is and stored in the receiving space of the buffer 203. Accordingly, the interface controller 204 checks the source address of the frame while deleting data frames stored in the receiving space. And, while continuing this confirmation, for at least any one (or all) of the addresses 91 that are all marked as'preliminary communication path in use', the data frame as the source address is a preset recovery time (for example, Check whether it is receiving more than a few minutes).

And, if a data frame with one (or all) addresses as source addresses among the addresses assigned to be relayed by itself is not continuously received for longer than the recovery time, the server related to the address recovers. This means that the spare bus is not used for more than a period of time, and this means that the switch to which the server (or servers) is connected has recovered from a failure and performs a normal relay operation, so that the use of the main communication path is resumed. 204 notifies the bus controller 202 of a'relay enabled state'. In addition, it updates its own relay allocation map accordingly.

The bus controller 202 notified of the'relay available state', according to any one of the above-described embodiments, transmits a live signal through a beacon field or a bit indicating'relay available' through a relay information notification field. By notifying other gate adapters of the value, other gate adapters also update their relay assignment maps accordingly, refer to the updated relay assignment map information, and switch to the data frames of the addresses currently assigned to them. Will be relayed.

In an embodiment according to the present invention, the gate adapter is implemented as an independent device, so that it is not connected to the switch through a communication line to which a specific communication protocol is applied, and the configuration of the gate adapter is included in the switch and integrated with the switch. It can also be configured. In this embodiment, the interface controller as described above may be configured to exchange signals according to a signal transmission/reception method of the switch internal bus or in a manner promised with the internal controller of the switch. And, in this embodiment, instead of the above-described method of indirectly grasping the relay failure state of the switch through the source address of data frames received through the spare bus, the interface controller may be configured to directly identify the operation failure of the switch. have. For example, an internal component of the switch that controls the switching operation of the switch or monitors the switching state is provided in the interface controller, and the recorded value is discharged after a certain period of time. For example, 1 can be recorded repeatedly at specified intervals. Then, the interface controller reads the value of the monitoring port at intervals longer than the predetermined time, and if the value indicates a discharged state, it determines that a failure has occurred in the relay operation of the switch, and is self-contained in the same manner as described above. The relay assignment map of is updated, and other gate adapters are notified of the'relay disabled status' through the bus controller.

If the value of the monitoring port, which is continuously checked afterwards, returns to 1 again, the interface controller notifies the other gate adapters of the'relayable status' as described above, so that the relay assignment map of all the gate adapters connected to the spare bus is displayed. It is updated according to the recovery of the corresponding switch.

The description of various embodiments so far has been made on the premise that an adapter, that is, a gate adapter, is connected to a down port or integrally provided for each communication device having a switching function. However, it is not necessary to have a 1:1 gate adapter. For example, as illustrated in FIG. 10, one gate adapter may be provided for each of two switches to secure a spare communication path. This configuration of the server system is suitable for application when the number of down ports of a switch provided to connect servers is small. The number of switches to be equipped with a gate adapter depends on the possibility of a switch failure or the amount of data the servers process to provide data service.N _S :N _G (N _S is the number of switches, N _G is Gate adapters may be provided for switches in a ratio of the number of gate adapters, N _S >N _G , and N _S and N _{G being an arbitrary number).} Of course, it is also possible to provide a gate adapter with a different ratio for each group classified according to the type of service provided to the servers.

Various embodiments of a system capable of providing a redundant communication path to a plurality of servers through a shared bus according to the present invention and the configuration and operation described in the embodiments are compatible with each other. Unless this is not possible, it can be selectively combined and implemented in various ways.

Above, the above-described embodiments of the present invention are disclosed for purposes of illustration, and those skilled in the art, within the technical spirit and scope of the present invention disclosed in the appended claims below, improve and change other various embodiments. , Substitution or addition may be possible.

1: main network line 2: spare network line
3 _i : main switch 4 _i : spare switch
10 _i ,20 _i : Server 11 _i : Main switch
21 _i : terminal adapter 22 _i : gate adapter
100,101: spare bus 201: transceiver
202: bus controller 203: buffer
204: interface controller

Claims (9)

In a server system for processing a request for a data service from client devices,
Each connection port consisting of a plurality of first ports and one or more second ports to which a line from the network is connected, and for data frames input from the connection ports, is assigned to the destination address of the data frame. Accordingly, N (N is 2 or more) communication devices configured to each perform a switching function that outputs to any one connection port,
The N each consisting of servers individually configured to transmit and receive data frames through a communication line, each directly connected to at least a part of the plurality of first ports provided in each of the N communication devices. Server groups,
Adapters of a first type provided in each of the servers belonging to the N server groups,
A second type of first adapter and a second adapter configured to be respectively connected to at least two of the N communication devices,
The adapters of the first type, the first adapter, and the second adapter are respectively connected to a spare bus, and the first adapter and the second adapter are selected from among servers belonging to the N server groups. The server receives the data frame transmitted to the spare bus through the adapter of the type 1 provided in it, and provides the data frame to the communication device to which it is connected among the N communication devices, by the switching function. For data frames received from the spare bus, data frames having a source address that matches any one of the addresses assigned to relay are accessed by themselves. It is configured to provide it to the communication equipment that has been prepared, and to discard the data frame that does not.
Each of the adapters of the second type connected to the spare bus, including the first adapter and the second adapter, relays all source addresses of data frames that can be received through the spare bus. Each relay assignment map assigned to be in charge is set,
The relay assignment map designates an order in which the adapters of the second type are in charge of relaying for all source addresses,
Each of the second type of adapters is assigned to a source address of an arbitrary data frame received from the spare bus, even if the order in which the relay assignment map assigns itself to the relay is not the earliest order. When an arbitrary adapter of the second type, which is assigned as a relay in charge of an order prior to the specified order, is designated in the relay assignment map because it is unable to relay, the arbitrary data frame is provided to the communication device to which it is connected. Server system. delete The method of claim 1,
Each of the second type of adapters detects a state in which the communication device to which it is connected does not normally perform the switching function, or detects that the communication line connection with the communication device to which it is connected is poor. Then, the server system configured to notify another adapter of the second type through the spare bus of its relay failure. In a server system for processing a request for a data service from client devices,
Each connection port consisting of a plurality of first ports and one or more second ports to which a line from the network is connected, and for data frames input from the connection ports, is assigned to the destination address of the corresponding data frame. Accordingly, N (N is 2 or more) communication devices configured to each perform a switching function that outputs to any one connection port,
Servers individually configured to be directly connected to at least a portion of the plurality of first ports provided in any of the N communication devices through a communication line to transmit and receive data frames through the communication line. Each of the N server groups,
Adapters of a first type provided in each of the servers belonging to the N server groups,
A second type of first adapter and a second adapter configured to be respectively connected to at least two of the N communication devices,
The adapters of the first type, the first adapter, and the second adapter are respectively connected to a spare bus, and the first adapter and the second adapter are selected from among servers belonging to the N server groups. The server receives the data frame transmitted to the spare bus through the adapter of the first type provided to the server, and provides the data frame to the communication device to which it is connected among the N communication devices, by the switching function. It is output to any one of the connection ports, and for data frames received from the spare bus, a data frame having a source address that matches any one of the addresses assigned to it is accessed by itself. It is configured to provide the communication equipment that has been used, and to discard data frames that are not
Each of the second type of adapters connected to the spare bus, including the first adapter and the second adapter, is a data cycle that is repeatedly formed on the spare bus. The identified random adapter on a relay assignment map that identifies the second type of random adapter that cannot transmit a signal, and allocates to relay all source addresses of data frames that can be received through the spare bus. A server system configured to designate relay-disabled for. The method of claim 4,
If the identified arbitrary adapter is an adapter that has formed the data cycle on the spare bus, among the second type of adapters, the second type of adapter designated to transmit the specific signal in the next section of the specific section. The server system, wherein a third adapter is configured to form a data cycle on the spare bus. delete In a server system for processing a request for a data service from client devices,
A plurality of first ports and connection ports consisting of one or more second ports to which a line from the network side is connected, and data frames input from the connection ports are Communication equipment configured to perform a switching function outputting to one connection port, and
A first group of servers individually configured to be directly connected to at least some of the plurality of first ports through a communication line to transmit and receive data frames through the communication line,
Adapters of a first type provided in each of the servers of the first group,
A second type of first adapter configured to be connected to the communication equipment,
Consisting of including a first spare bus and a second spare bus,
The adapters of the first type and the first adapter are respectively connected to the first spare bus and the second spare bus so that one of the spare buses can be selectively used,
The first adapter receives a data frame transmitted to the first spare bus or the second spare bus through the first type of adapter provided in the server of the first group, and the By providing the communication equipment, the server system is configured to be output to any one of the connection ports by the switching function. The method of claim 7,
Each of the first type of adapters and the first adapter checks the load levels of the first and second spare buses, respectively, and selects a spare bus having a lower level of the checked load, and a data frame Server system configured to be used for transmission of. The method of claim 7,
Wherein the first type of adapters and the first adapter are configured to select a spare bus that can be used first among the first spare bus and the second spare bus, respectively, and use them for transmission of a data frame.

Images