JPH023845A - Communication protocol processing system - Google Patents

Abstract

PURPOSE:To easily make a communication protocol processing into the distributed processing of inter-cluster and to recognize an opposite system by connecting first and second communication protocol processing functions in respective computers by means of the functions belonging to the same or different computers. CONSTITUTION:The first communication protocol processing functions A-1 to A-3 and the second communication protocol processing functions B-1 to B-3 in respective computers #1-#3 are not limited by the connection of elements in respective computers, but connect the first or second communication protocol processing functions in the optional computer and they can be used for a communication processing with one opposite system. Application processing parts C-1 to C-3 are not limited similarly by the frames of the computers, but they can be distributed among the optional computers. Consequently, parallelism which is more rough than the units of opposite systems 2-4 can be shared by plural clusters 5-7, and the opposite system 2-4 can be caused to deal the self system 1 as a single system.

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] The present invention relates to a communication protocol processing method for a system in which a plurality of computers are loosely coupled to perform communication processing with a partner system via a network.

The purpose is to facilitate the distributed processing of communication protocol processing between clusters in a system that includes multiple computers as a cluster, and to enable recognition of partner systems independent of the route between the systems.

The communication protocol processing function of each computer is a first communication protocol processing function that performs communication processing for each partner system without depending on one communication line, and a second communication protocol processing function that performs communication processing that depends on one communication line. Separately set up and
The first and second communication protocol processing functions in each computer are
It has a configuration that allows items belonging to the same or different computers to be combined.

[Industrial application field]

The present invention relates to a communication protocol processing method for a system in which a plurality of computers are loosely coupled to perform communication processing with a partner system via a network. To provide a communication protocol processing method that enables communication between systems to be executed without making the other system aware of it.

[Conventional technology]

When a system in which multiple computers that do not share main memory are loosely coupled as a cluster communicates data with other systems via an OSI network, for example,
It is desirable to operate as one consistent system without making the other system aware of individual computers.

However, as shown in Figure 8, when multiple routes exist between each computer in the system and the other party's system, conventionally communication processing is performed with each individual computer or route in mind. From the system's point of view, it was not possible to maintain the integrity of the system.

FIG. 9 shows a system consisting of a plurality of loosely coupled computers (ie, a cluster) and an example of distributed communication processing.

In FIG. 9, 10 is one system, 11 to 1
3 is the computer #l of each cluster.

#2. #3.14 is an inter-cluster communication path, A-1, A-2, and A-3 in each computer are application processing units that require communication as a means of communication, and C-1, C-2 and C-3 are This is a communication protocol processing unit.

In the illustrated conventional system, an application processing unit (A-n) in each computer shares communication processing with a corresponding communication protocol processing unit (C-n) in the same computer.

In the case of this method, the following operational problems arise.

Rigidization of processing allocation The computer to which application processing that uses communication is allocated is limited by the location of the line used.

This also makes it difficult to adjust the load balance.

Limitation of communication range Once the computer that processes the application is determined, the range of routes that can be used for communication is conversely limited to lines belonging to that computer.

FIG. 1θ shows an example of distributed processing using another conventional method.

In the example shown in Figure 10, the application processing unit and the communication protocol processing unit responsible for its communication processing do not need to be in the same computer, and the communication processing of the application processing unit in each computer is routed every 9 communications. is assigned to the computer to which the communication line is set.

For example, when the application processing unit of a certain computer uses the communication line in computer #l, it must request communication processing to the communication protocol processing unit A-1 of computer #l, no matter which computer it is from. . Computing engine communication in this case is performed using a computing engine communication path.

However, in the case of the conventional example shown in FIG. 8, the following problem arises in the implementation of one communication protocol processing section.

One piece of data is divided by network (N) 75,
When these are multiplexed and transmitted using a plurality of communication lines belonging to different computers, there is no place to assemble the received data (network protocol data units).

When communicating by connecting multiple circuits in the transport (T) layer, there is no place to merge received data (transport protocol data units).

Processing that needs to be performed in synchronization with each partner system cannot be completed on a single computer, making processing complicated. For example, when one communication protocol process is distributed to multiple computers for parallel processing as shown in Figure 11, extra synchronization processing is required between each computer to number sequential data. be done.

[Problem to be solved by the invention]

In conventional communication protocol processing in a system including multiple computers as a cluster, it has not been easy to improve processing capacity and fault tolerance by performing distributed processing among multiple computers.

Furthermore, when there are multiple routes that can be reached between the system and the other system, it is not possible to make the other system recognize the system as a single system regardless of the routes.

The present invention is intended to facilitate distributed processing of communication protocol processing between clusters in a system including a plurality of computers as a cluster, and to enable recognition of partner systems independent of routes between systems.

[Means to solve the problem]

The present invention divides the two communication protocol processing into communication protocol processing for each partner system and communication protocol processing that depends on the communication line connected to the cluster computers,
The above-mentioned problem is solved by enabling distributed processing using separate clusters.

FIG. 1 is an explanatory diagram of the principle of the present invention. In fig.

1 is the own system.

2.3.4 are the partner systems X and Y, respectively.

It is Z.

5.6.7 are computers #1, #2, in clusters that are loosely coupled and do not coexist with the main host jlJ in their own system 1. #
It is 3.

8 is computer #1. #2. This is an inter-cluster communication path that loosely couples #3.

9 is one communication line.

A-1, A-2, and A-3 are each computer #1. ＃2゜＃3
This is the first communication protocol processing function that performs communication processing for each partner system within the system.

B-1, 8-2, and B-3 are each computer #1. #2, #3
This is a second communication protocol processing function that performs communication processing depending on the communication line 9 connected to each computer within the computer.

C-1, C-2, and C-3 are each computer #1. ＃2゜＃3
This is the application processing section within.

Each computer #1. #2. The first communication protocol processing functions A-1, A-2, A-3 and the second communication protocol processing functions B-1, 8-2, B-3 in #3 are elements in each of the five computers. The first or second communication protocol processing functions of any computer can be combined and used for communication processing with one partner system without being restricted by the connection between the two.

The assignment of the first or second communication protocol processing functions to the application processing units C-I C-2 and C-3 is not limited to the framework of two computers, and can be distributed between arbitrary computers.

[Effect]

In Figure 1, the communication connections between the systems are the first and second
maintained by a combination of communication protocol processing functions. The first and second communication protocol processing functions assigned to one route do not need to be in the same computer, but information between each function and the application processing unit when functions in different computers are combined. Communication is performed using an intercluster communication path 8. However, in the figure, these paths are shown directly using double lines for clarity.

At a certain point in time, the first communication protocol processing function for each partner system is assigned to only one computer. It may be assigned to a different computer at another time. In this way, the first communication protocol processing function of each partner system can be easily processed.

The second communication protocol processing function described above should always be assigned to a computer to which one line is connected, but if the adjacent system to which that line is connected changes (circuit-switched network (in the case of a subscriber line),
It may be necessary to switch the processing contents (protocol types and operating parameters of layers lower than the N layer) to ones that are compatible with the adjacent system.

In the example shown in FIG. 1, the application processing unit C-1 of computer #1 communicates with partner systems X and Z, and the application processing unit C-2 of computer #2 and the application processing unit of computer #3 communicate with each other. C-3 each communicates with the other party system Y.

At this time, the application processing unit C-1 uses A-1 of the first and second communication protocol processing functions for communication processing with the partner system X.
and B-1, and A-3 and B-3 are respectively assigned to communication processing with the partner system Z.

On the other hand, the application processing units C-2 and C-3 use the second communication protocol processing functions B-1 and B-2 to set two routes in parallel with the partner system Y and perform communication processing. ing.

FIG. 2 is an example of communication processing using the two routes mentioned above, when using OSt-compliant CLNP (Connectionless Network Protocol).

The partner system Y sends the divided NPDUs (network protocol data units) of N5DUs (network service data units) to computers #1 and #2 via separate routes.
This shows the case where the data is transmitted to In this case, each LAN
The second communication protocol processing function B-1 via the connection point of
,B-2 is divided into 1 (NPDU, ,NP
DU, .

This allows the system to uniquely recognize the other system regardless of the route (LAN connection point), and limits the connection points to the LAN that can be used for communication with the other system for each partner system. It is also no longer necessary to do so.

Also, by gathering the first communication protocol processing function for each partner system into one computer.

Processes that can be controlled exclusively within one computer and that need to be processed in synchronization with each partner system are controlled by N41. That is, if communication processing for each partner system is distributed among a plurality of computers as shown in FIG. 3, a local protocol is required to synchronize the computers, which complicates control.

Furthermore, by dividing and assigning the two communication processes to the first and second communication protocol processing functions, it becomes possible to determine a single processing location for the branch/merging process that handles a plurality of connections at once.

Also, by separating communication processing and application processing.

There are no restrictions between the location of lines used for communication and the placement of applied processing on computers.

〔Example〕

Next, an embodiment in which an O51 compliant protocol is applied as a communication protocol to the hardware of a loosely coupled cluster system will be described with reference to FIGS. 4 to 6.

Here, in a loosely coupled cluster system, multiple computers (clusters) that do not have main memory coexist can use a data transfer mechanism (intercluster path (ε path)) that is comparable to in-memory transfer.
A computer system connected by

FIG. 4 is a basic configuration diagram of a loosely coupled cluster system, in which 20 is a loosely coupled cluster system, 21 to 23 are clusters, 24 to 28 are adapters for lines and LAN, and 29 is an intercluster communication path. , FIG. 5 shows an example of distribution of communication protocol processing, and FIG. 6 shows an example of communication between clusters.

First, the main abbreviations used in the description of this embodiment are shown below.

CL N P Connection-Less Ne
twork Protocol COConnectio
n-OrientedL A N Local Ar
ea NetworkL A N A LAN A
daptorM A CMedia Access C
ontrolN CNetwork Connect
tionN CM S Network Connect
tion Management 5ubprotoc
ol N P D U Network Protocol
Data l1nitN S A P Network
5service Access Po1nLN S
D U Network 5service Data
Unit S N P A SubNeLwork Po
1nt of AttachmentT CTrans
port ConnectionT P D U Tr
ansporL Protocol Data Uni
tV CVirtual UP to C4rcuitO3+ protocol (1) Processing unit configuration a) Two secondary units are provided as units that can be processed in parallel (see Figure 5) ■ Communication protocol processing for “remote open system unit” - Tlli ( ) Communication protocol processing for layers higher than transport layer) is assigned here.

- Among the processes of the N layer (two-node layer), the received NPDU assembly process that handles CLNP is also included here. 5 NICs
Some of the P-roll is also included here.

■ Processing that depends on the line connected to the cluster. Among the processing that corresponds to the line and the part that connects to the LAN,
The parts that are not handled by adapters etc. are classified here. A part of the SN I CP role of the N layer, the 5NACP role, and the DLC layer are included here.

Individual connections are maintained through the processes of (1) and (2). There is no restriction that ■ and ■ are assigned to the same cluster for each connection. If necessary, use an inter-cluster path between ■ and ■ (see *1 in Figure 5).

b) At a certain point in time, the above-mentioned communication protocol processing for each "partner system" is assigned to one cluster (individual computers forming a loosely coupled cluster system). However, Y may be assigned to a different cluster at another time. - In this way, the processing of ■ is collected into one cluster to make it easier to process.

C) The above process (■) is always assigned to the cluster to which 9 lines are connected. However, if there is a change in which adjacent system the line is connected to (a line joining the line-switched knitting yarn network), the content of the processing will be changed to match the adjacent system.

d) CLNP's received NPDU assembly process.

Do this after grouping into units of ■. Therefore, if it is necessary to exchange NPDUs that may have been divided into one unit between the processing units of ■ and ■ across clusters,
Do this.

(2) Determining the arrangement of processing units An example of the arrangement of (2) and (2) will be explained starting with a simple (2).

a) Arrangement of the above process (2) The process (2) is fixed to the cluster to which the lines to be used (including adapters that share processing protocols below NIW, such as LAN and SDN) are connected.

b) A means for limiting the number of clusters in which the process (■) is located in the unit of arrangement partner system for the process (■) above to at most one is provided.

For example, there is a method of creating only one processing unit to make this decision (one form is a processing unit with decision-making authority), and there is also a method of making this decision through discussion among clusters (
There is also a form in which there are multiple processing units that have decision-making authority.

In this embodiment, this determination is made in three primary cases.

1) 3 from 5 application processing to establish upper connection
If there is a request somewhere in the part that processes communications.

If the unit (2), which processes communication with the partner system to which one application wants to connect, is already located in any cluster at that time, the current status is followed.

- If (1) does not yet exist and therefore the placement cluster has not been determined yet, the placement cluster is determined by the above-mentioned means. This judgment is.

For example, the decision may be made from the following input information:

- Specified in advance. Cluster placement for each partner system.

・Cluster where the application process that issued the communication request exists.

・Cluster that is convenient for processing connections below the N layer established with the other system (■1), that is, 1
A cluster in which network resources (lines) that are likely to be used for communication exist.

- Status of processing and load sharing for each cluster in the FCMP system.

2) When the NC (N15 connection: network connection) receives a call from the other party's system.

If the unit (2) that processes communication with the system at the other end of the incoming NC (destination system) is already placed in any cluster at that time, the current status is followed.

- If (2) does not yet exist and therefore the placement cluster has not been determined yet, the placement cluster (2) is newly determined. This judgment may be made from input information such as 9th order, for example.

- Specified in advance. Cluster placement for each partner system.

- A cluster for processing connections below the N layer that processes incoming NG calls (set as ■2).

・Processing sharing status and load sharing status of each cluster in a loosely coupled cluster system.

3) When receiving NPDU of CLNP.

- If the unit (2) that processes communication with the received NPDU sending system (other system) is already located in any cluster at that time, the current status is followed.

- If ■ does not yet exist, the same as 2) above.

(3) Simplified protocol processing By gathering the processing of ■ into at most one cluster, many of the processes that need to be processed in synchronization for each partner system can be closed within the cluster as shown in the first order. Process.

a) TC reference management processing of Tli protocol (including NCMS) Zero-order processing is performed in one exclusive unit in one cluster for each partner system. Since there is one exclusive unit, it can be easily realized.

- Demultiplexing processing at the T layer (based on the local TC reference value carried in the received TPDU).

(Selection of target TC) - New assignment of TC reference on the 0 station side (Numbering to avoid duplication) Termination of use and freezing processing of the TC reference on the 1 station side (management of timing when it can be reused) - TC reference on the other side Duplicate check - New assignment of local NC reference (numbering to avoid duplication) - Termination of use and freezing of first station NC reference (management of timing when it can be reused) - Duplication of other side NC reference Inspection b) Processing using NC in Tli (TN class 4 branch/merging processing) Even if multiple NCs are used for each partner system, the T layer processing using those NCs can be performed in one Processing is performed in one exclusive unit within the cluster. This is because the exclusive unit can be combined into one in this way.

The subsequent branching/merging process of the T-layer lath 4 can be easily realized.

- If necessary, freeze TPDUs that are to be sent on different TCs to the same partner system.

Divides the flow to different NCs.

- N5 received from different NCs to the same partner system
The DU is separated into TPDUs and merged into the TC corresponding to each TPDU.

c) CLNP NPDU assembly processing When CLNP is used as the NN protocol and N5DO division/assembly is performed, the processing is performed as follows.

1. From the address of the source system included in the received NPDU, the processing unit (3) for that process is determined ((3) in (2) above), and the received NPDU is sent to (3). Even if the processing units (■) that received individual NPDUs are not in the same cluster, they are collected into one processing unit (■□) through this process.

-■ What is the process of assembling the NPDUs collected in 3?

This is done in one exclusive unit within the cluster. As a result of assembly, one entire N5DU is obtained as a cluster in which δ+1 is assigned to ■.

(4) Comparison of inter-cluster communication δ with the conventional method A comparison between the inter-cluster communication in the embodiment of the present invention and the inter-cluster communication in the conventional method explained in FIG. 10 is made using FIGS. 6 and 7. explain.

Note that it is assumed that the shaving of the processing unit (2) is performed in a concentrated manner.

FIG. 6 is an explanatory diagram of inter-cluster communication in the embodiment of the present invention. In the figure, ■ and ■ indicate inter-cluster communication, and each has the following contents.

■ Exchange for centrally managing the processing location of each partner system.

■ The following flow of data and control within the N layer that connects the processing unit ■ (unit of the other system) and the processing unit ■ (processing dependent on lines, etc.).

- Data and control flow within the N layer for N layer relaying.

- CLNP's N P D U i, [I decided on a location in Mitate.

The process of collecting received NPDUs there.

- Exchange for centrally managing circuit-switched subscriber lines.

Next, FIG. 7 is an explanatory diagram of inter-cluster communication in the conventional method, and .circle.--. . . .-. in figure 2 indicate inter-cluster communication, each of which has the following contents.

■ Data (in T PDU units) and control flow within the T layer for the shunt function of the T layer lath 4.

■ Exchange for centrally managing reference operations within the T layer.

■ Exchange for central management of activity identifiers.

■ The following exchanges are used to determine the termination of the called NC and to transfer data within the N layer.

- Data and control flow within the NJW for NJW relaying.

-Determine the location for CLNP NPDU assembly.

The process of collecting received NPDUs there.

- Exchange for centrally managing circuit-switched subscriber lines.

Here, ■ in FIG. 6 and ■ in FIG. 7 have the same content.

〔Effect of the invention〕

According to the present invention, it is possible to share parallelism that is rougher than that of the other system among multiple clusters, and even if there are multiple routes between the other system and the other system, the own system can be connected to the other system in a single manner. can be treated as a system.

Furthermore, when deciding which cluster in the system should be assigned communication protocol processing with which partner system,
It is easier to take into account the relatively fixed relationships that typically exist between network equipment and clusters, making it easier to efficiently utilize equipment and maintain performance.

Furthermore, work that requires consistency in each partner system can be processed within one cluster, which has the advantage of simplifying inter-cluster communication.

[Brief explanation of the drawing]

Figure 1 is a diagram explaining the principle of the present invention, Figures 2 and 3 are diagrams explaining the operation of the present invention, Figure 4 is a basic configuration diagram of a loosely coupled cluster system according to an embodiment of the present invention, and Figure 5 is a communication protocol. FIG. 6 is an explanatory diagram of inter-cluster communication in an embodiment of the present invention, FIG. 7 is an explanatory diagram of inter-cluster communication in the conventional method, and FIG. 8 is an explanatory diagram of an example of processing distribution. The explanatory diagrams, FIGS. 9 and 10, are explanatory diagrams of a conventional method of distributed processing within the system, and FIG. 11 is an explanatory diagram of distributed processing of communication protocol processing. 1 in Figure 1: Own system 2-4: Opposite system X, Y, Z 5-7: Computer (cluster) #1. #2. #38: 1ffl communication path between clusters: communication line or LAN

Claims (3)

[Claims] (1) Consisting of multiple loosely coupled computers (5, 6, 7), each computer (5, 6, 7) is equipped with a communication protocol processing function and communicates with the other system (2, 3, 4). In the system (1), the communication protocol processing function of each of the computers (5, 6, 7) is combined with a first system that performs communication processing for each partner system without depending on the communication line. Communication protocol processing function (A-1 to A-
3) and a second communication protocol processing function (B-1 to B-3) that performs communication processing depending on the communication line. A communication protocol processing method characterized in that the second communication protocol processing functions (A-1 to A-3, B-1 to B-3) are capable of connecting those belonging to the same or different computers. (2) The communication protocol processing method according to claim 1, wherein only one first communication protocol processing function (A-1 to A-3) is assigned to each partner system. (3) In claims 1 and 2, the first and second communication protocol processing functions (A-1 to A-3, B-1 to B-3)
The communication protocol processing method is characterized in that the application processing units (C-1 to C-3) that request communication processing can be connected to those belonging to the same or different computers.