JPH07219902A - Network adapter and network system - Google Patents

Abstract

(57) [Summary] (Modified) [Purpose] To reduce the broadcast reception processing load on the main processor of a computer in a system that matches memory contents by periodic broadcast. [Structure] A computer 201 includes a main processor 301 and a main memory 401, and a network adapter 202 is connected to a system bus 402. The memory transfer area 302 is provided on the main memory 401, and the local memory 404 is provided with the soft switch table 801. In addition, the local processor 405 and the transmission path control LSI 40 are also mounted on the adapter.
8. There are local buses 407. Local processor 4
Reference numeral 05 is activated by an interrupt from the transmission path control LSI 408 and determines whether or not the received broadcast data is data to be captured by the own computer, by referring to the soft switch table 801. If the data is to be fetched, the main processor 301 is notified, otherwise, it is discarded.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a network adapter and a network system, and more particularly to a network capable of reducing the load of the matching process of the cycle transfer memory for sharing information among a plurality of computers connected to the network. The present invention relates to adapters and network systems.

[0002]

2. Description of the Related Art Conventionally, in a network system to which a plurality of computers are connected, a cyclic (cyclic) memory transfer method has been used as one method for sharing data. In this method, each computer has a transfer memory for sharing data.
Each computer is connected to the network via a network adapter, but the transfer memory is provided in the local memory of the network adapter.

For example, when data is shared by three computers # 1, # 2 and # 3, a transfer memory is provided in the network adapter of each computer. The transfer memory is provided with an area for computer # 1, an area for computer # 2, and an area for computer # 3. Then, the network adapter of the computer # 1 broadcasts the data of the computer # 1 area of the transfer memory by the local processor on the adapter. In the network adapters of the computers # 2 and # 3 that received the broadcast data, the local processor on the adapter stores the data in the computer # 1 area of the transfer memory. As a result, the data in the computer # 1 area will be shared. The same applies to the area for computer # 2 and the area for computer # 3.

FIG. 17 is a diagram showing the internal hardware configuration of the computer 201 and the network adapter 403 of the conventional form used for executing the memory transfer, and the connection relationship between the computer 201 and the network adapter 403. A plurality of computers and network adapters configured as shown in this figure are connected to the network.

The computer 201 includes a main processor 301 and
It includes a main memory 401 and a system bus 402. The main processor 301, main memory 401, and network adapter 403 are connected to the system bus 402.

The network adapter 403 comprises a transmission path control LSI 408, a local memory 404, and a local processor 405. Transmission line control LSI 40
Reference numeral 8 controls the interface with the network 203.
The local memory 404 is provided with a memory transfer area 302 and a buffer area 406. The buffer area 406 is for temporarily storing transmitted / received data. The memory transfer area 302 is for sharing data with other computers. Local processor 4
Reference numeral 05 controls the network adapter 403.

As shown in FIG. 17, the conventional memory transfer area 302 is on the local memory 404 in the network adapter 403, and the main processor 301 of the computer 201 accesses the local memory 404. It is the responsibility of the local processor 405 to periodically broadcast the data in the memory transfer area 302.

FIG. 18 is a diagram showing the distribution of communication processing between the computer main processor 301 and the local processor 405 of the network adapter 403 in the conventional configuration.

In addition to the memory transfer processing described so far, the communication processing includes normal data communication processing for message passing. Conventionally, the main processor 301 on the computer main body executes application-level processing for both normal data and memory transfer.
The local processor 405 on the network adapter 403 executes the 2 to 4 layers in the 7-layer model of OSI, that is, the MAC sublayer, the LLC sublayer, the network layer, the transport layer, and the memory transfer cycle activation. .

FIG. 19 is a diagram showing the relationship between the position of the memory transfer area 302 and the main processor 301 in the conventional memory transfer system.

Main processor 3 of computer # 1 (201X)
01X writes data to the computer # 1 area 303X of the memory transfer area 302X on the conventional network adapter 403X. Data in the computer # 1 area 303X is periodically broadcast using the network 203. It is the local processor 405 on the network adapter 403 that performs this periodic broadcast.

Network adapters 403Y, 403Z
Receives this broadcast data, and the computer transfer area 30 of the memory transfer areas 302Y and 302Z
Data is written in 3Y and 303Z. Calculator # 2 (20
1Y), # 3 (201Z) main processor 301Y, 3
01Z are network adapters 403Y and 403, respectively.
Computer for memory transfer areas 302Y and 302Z on Z #
By reading the 1st areas 303Y and 303Z, the data written by the computer # 1 (201X) is shared.

FIG. 20 is a diagram showing a frame format for broadcasting memory transfer data.

A tag 702 is attached to the memory transfer data 701 to form memory transfer information 703. On the network transmission path, it is transmitted as the broadcast data frame 304 in which the MAC header 704 is added to the memory transfer information 703.

As described above, in the conventional periodic memory transfer system, the memory transfer area is provided in the local memory on the network adapter, and the periodic broadcast data transmission / reception processing for matching the memories is performed by the local processor on the adapter. It was realized by using a dedicated network adapter that is executed in.

[0016]

However, the openness of network systems is advancing at the present time, and many computers are connected to the network by general-purpose network adapters. An approach to construct a periodic memory transfer system using a general-purpose network adapter instead of using it is also under consideration.

A general-purpose network adapter is, for example, one described in "intel: iSBC 186/51 Data Sheet". The computer-adapter interface of a general-purpose network adapter often adopts a system having a shared memory that can be accessed by both the local processor and the computer main processor. However, its capacity is limited, and it is difficult to put the memory transfer area on the shared memory. Also, the local processor is
The transmission path control LSI is driven and all the received data is notified to the computer main processor. Therefore, so-called communication protocol processing is performed by the main processor.

In the technique using such a general-purpose network adapter, the conventional message passing is
In many cases, two-to-peer communication was used, and when a broadcast-based system was rare, it was effective without increasing the load on the computer main processor.

However, the use of the above-mentioned general-purpose network adapter in the system for realizing the periodic memory transfer based on the broadcast has the following problems.

(A) Since memory matching is achieved by broadcasting, if memory transfer is executed in a short period,
The load of the broadcast data reception process increases.
Therefore, a computer with a low performance of the main processor of the computer cannot join the memory transfer system because the load of communication processing is large and the original application processing is hindered.

(B) Not all computers connected to the network perform memory transfer. But,
Because it is a broadcast, unrelated data is also frequently received, wasting the power of the computer main processor.

It is an object of the present invention to provide a method for reducing the load of reception interrupt of a computer main processor when memory transfer is realized by using a general-purpose network adapter.

Still another object of the present invention is to provide a system which does not impose an extra communication processing load on a computer not participating in the memory transfer by receiving broadcast data for matching the transfer memories.

Still another object of the present invention is to provide a network adapter having the same hardware / software configuration and capable of varying the interrupt cycle to the main processor by means of a switch.

Still another object of the present invention is to realize data sharing by memory transfer between computers having different main processor performances.

[0026]

In the network adapter and the network system according to the present invention, the execution of the process of thinning the broadcast data is added to the local processor of the network adapter which was executing the drive process of the transmission path control LSI. As a thinning method,
For example, regarding the received broadcast data, the broadcast cycle is determined, and when the cycle is the cycle of broadcast data to be captured by the computer, the received data is notified to the processor of the computer,
If not, the received data is discarded. Alternatively, the number of times the broadcast data is received may be counted, the received data may be notified to the processor of the own computer only when the number of times reaches a predetermined number, and other received data may be discarded.

Further, the local processor on the network adapter of the computer not participating in the memory transfer discards the broadcast data for making the transfer memory consistent.

Further, it is preferable to provide a switch for setting information for determining whether to notify the received broadcast data to the processor of the own computer or to discard it without notifying it. The switch may be a hardware switch or a soft switch table that can be downloaded from the computer side to the network adapter.
The information set by the switch is, for example, the value of the cycle when the above determination is made in the broadcast cycle, and the number of times when the above determination is made by the number of times of reception.

Further, in the network system according to the present invention, the network adapter has a variable interrupt period for computers having different main processor performances.

[0030]

The process of thinning out broadcast data in a network adapter is a process of notifying the computer of the broadcast data to be captured by the computer (local computer) to which the network adapter is connected, and discarding the other data. For example, in the case of making the determination based on the broadcast cycle, when the memory transfer data other than the cycle to which the computer is subscribed is received, the data is discarded and the main processor is not notified. In the prior art, the main processor of the computer had no choice but to perform reception processing even for data unrelated to its own computer for broadcasting, but in the present invention, it thins out only the necessary amount at the network adapter level. It is possible to reduce the load related to the matching processing of the cycle transfer memory of the main processor of the computer.

Since the network adapter of the computer not participating in the memory transfer discards all the broadcast data and does not notify the main processor, the communication processing load of the main processor, especially the processing load of receiving the broadcast data can be reduced.

The local processor on the network adapter judges whether or not to notify the main processor of the connected computer of the reception of broadcast data by looking at the switch. This switch (hardware switch or soft switch table) can be set by the user.

If this switch is composed of a soft switch table, setting information can be downloaded from the connection computer and set when the adapter is initialized. In particular, when a group that shares data is defined based on the performance of the computer (particularly the main processor), setting information suitable for the performance of the connected computer can be set in the soft switch table. For example, when a group of high-performance computers broadcast in a short cycle, and a group of low-performance computers broadcast in a long cycle, when initializing the adapter,
It is possible to download the information of the cycle corresponding to the performance of the computer from each computer and set it in the soft switch table. In this way, by using a network adapter with a variable interrupt cycle, broadcast data can be thinned out so that the load on the main processor does not become high, so even if a memory transfer system is constructed between computers with performance differences, It is not necessary to adjust the cycle to a computer with a low.

[0034]

Embodiments of the present invention will be described below with reference to the drawings.

[First Embodiment] FIG. 2 shows a configuration example of a network system using a network adapter according to an embodiment of the present invention.

This network system includes a computer 20.
An end system including 1A and a network adapter 202A, a computer 201B, and a network adapter 2
02B, an end system including a computer 201C and a network adapter 202C, a computer 201D and a network adapter 202D
And an end system including a computer 201E and a network adapter 202E,
An end system including a computer 201F and a network adapter 202F is connected to the network 203.

In FIG. 2, computers 201A, 201
B and 201C have relatively high computer main processor performance, and computers 201D, 201E and 201F have relatively low computer main processor performance.

FIG. 3 is a diagram showing the concept of periodic (cyclic) memory transfer implemented in the network system as shown in FIG. Cyclic memory transfer is one of the methods to share data among multiple computers, and the data in the local computer area of the transfer memory area is periodically broadcast to other computers to transfer the data to the transfer memory area. Match the data of.

In FIG. 3, computer # 1 (201X) having the main processor 301X and transfer memory 302X, and computer # 2 having the main processor 301Y and transfer memory 302Y.
(201Y), main processor 301Z and transfer memory 3
An example of performing memory transfer with a computer # 3 (201Z) having 02Z will be shown.

The transfer memory 302 is for each computer # 1, #
A fixed area is allocated to the # 2 and # 3 areas, and the main processor 301X of the computer # 1 (201X) can read and write the # 1 area 303X, and the # 2 area and the # 3 area. Can only be read. Data in the # 1 area 303X is periodically transmitted as broadcast data 304 to the computer # 2.
(201Y) and computer # 3 (201Z). Computer # 2 (201Y) and Computer # 3 (201
Each of the # 1 areas 303Y and 303Z in (Z) is rewritten to the same value as the area 303X. Calculator # 2, #
The No. 3 can share the same data as the computer # 1 by reading the values of the # 1 areas 303Y and 303Z. Similarly, for the area for # 2, computer # 2 (201Y)
However, for the area for # 3, computer # 3 (201Z)
However, by broadcasting each, the transfer memories are matched.

FIG. 4 is a diagram showing the internal hardware configuration of the computer 201 and the network adapter 202 in the network system of this embodiment shown in FIG. 2 and the connection relationship between the computer 201 and the network adapter 202.

The computer 201 includes a main processor 301, a main memory 401, and a system bus 402, and a network adapter 202 is connected to the system bus 402, which is the same as the conventional example shown in FIG. The difference is that
The memory transfer area 302 is provided on the main memory 401, and the soft switch table 801 is provided on the local memory 404 on the network adapter 202. On the network adapter 202, the local processor 405, network 20
3, a transmission path control LSI 408 that controls the interface with the
And a local bus 407 to which they are connected.

The soft switch table 801 is a table for storing information as a reference for determining whether or not to notify the main processor 301 of the own computer 201 when the broadcast data is received from another computer. The data of the soft switch table 801 is downloaded from the connection computer and set when the network adapter is initialized.

In this embodiment, two types of memory transfer broadcast periods, long and short, are provided according to the performance of the main processor. They are 1 ms cycle and 4 ms cycle. The computer belonging to the 1 ms cycle group broadcasts the memory transfer in the 1 ms cycle, and the computer belonging to the 4 ms cycle group broadcasts the memory transfer in the 4 ms cycle. The soft switch table 801 contains the cycle (1 ms or 4
ms) is stored. The local processor 405 of each network adapter uses the soft switch table 8
By comparing the value of the cycle stored in 01 with the value of the cycle in the received broadcast data, it is possible to determine whether to take in or discard the broadcast data.

FIG. 5 is a diagram showing the distribution of communication processing between the main processor 301 of the computer 201 and the local processor 405 of the network adapter 202 in this embodiment.

In the main processor 301 on the computer side,
The LLC sub-layer of normal data, the network layer, the transport layer, and the processing of the application, the periodic start of memory transfer, and the processing of the application are executed. Local processor 40 on the network adapter side
At 5, for MAC sublayers, especially memory transfers, only broadcast is performed.

FIG. 6 is a diagram showing the position of the memory transfer area and the relationship with the main processor in the memory transfer system of this embodiment.

Main processor 3 of computer # 1 (201X)
01X is the # of the memory transfer area 302X on the main memory.
The data of the 1st area 303X is broadcast using the network adapter 202 and the network 203. Computer # 2 (201Y), # 3 (201Z)
The main processors 301Y and 301Z of the respective CPUs transfer the data received via the network adapters to the # 1 areas 303Y and 303Z of the memory transfer areas 302Y and 302Z.
To store. Then, computer # 2 (201Y), # 3
The main processors 301Y, 301Z of (201Z) are the # 1 areas 30 of the memory transfer areas 302Y, 302Z.
By reading 3Y and 303Z, computer # 1
Share the written data of (201X).

FIG. 7 is an example of a system configuration in which, in this embodiment, two types of memory transfer cycles are provided in accordance with the main processor performance.

As shown in FIG. 2, computer A (201
A), computer B (201B), and computer C (201
In the case of C), since the main processor performance is high, it belongs to the 1 ms period group, and broadcasts at 1 ms period to share data. On the other hand, computer D (201
D), computer E (201E), and computer F (201
In the case of F), since the main processor performance is low, it belongs to the 4 ms cycle group, and broadcasts are executed at the 4 ms cycle to share data.

However, since all of these computers are connected to the same network 203, the conventional method has a problem that the load of reception processing by broadcast data becomes the same as that of the 1 ms period group even for the 4 ms period group. Therefore, in the computer of the 4 ms cycle group, the local processor on the network adapter discards the broadcast data for the 1 ms cycle to reduce the broadcast data reception processing load of the main processor of the computer main body.

FIG. 1 is a flow chart showing the procedure of reception processing in the local processor 405 on the network adapter 202 of this embodiment.

The local processor 405 is activated by a reception interrupt from the transmission path control LSI 408, and first the received data is the broadcast data 304 (FIG. 2).
It is determined whether it is 0) (step 101). If it is not the broadcast data 304, the data reception is notified to the main processor 301 as it is (step 105). In the case of broadcast data 304, the soft switch table 801 is referred to check the cycle to which the computer belongs (step 102), and the memory area tag 702 of the broadcast data 304 is analyzed (step 103). If it is the data of the subscription period, the main processor 301 is notified, otherwise the broadcast data is discarded (step 104).

The frame format for broadcasting the memory transfer data is the same as in FIG. 20 described in the conventional example. However, the memory area tag 70
It is assumed that 2 stores the broadcast cycle (1 ms or 4 ms) of the broadcast data. This is referred to in step 103 above.

FIG. 8 is a flow chart showing the procedure of the receiving process in the main processor 301 of the computer 201 of this embodiment.

The main processor 301 receives a notification that there is received data by an interrupt or the like from the local processor 405 of the network adapter 202. Upon receiving the notification, the received data is transferred to the memory transfer information 703 (see FIG. 2).
0) or not (step 1201), if the memory transfer information 703 is stored in the memory transfer area 302 (step 1202), and if it is not the memory transfer information 703, it is processed as normal data (step 1203).

FIG. 9 is a flow chart showing the processing procedure in the main processor 301 of the computer 201 for transmission.

The main processor 301 of the computer 201 causes the memory transfer data 701 to be activated by the periodic activation by the timer.
After writing (FIG. 20) to the transmission buffer 406 and adding a memory area tag (step 1301), the network adapter 202 is activated. The computer 201 in the 1 ms period group performs the periodic activation by the timer in the 1 ms period, and the computer 201 in the 4 ms period group performs the 4 ms period. Further, the memory area tag to be added includes the cycle of the group to which the computer belongs.

FIG. 10 is a flow chart showing the procedure of transmission processing of the local processor 405 on the network adapter 202.

When the main processor 301 of the computer 201 is activated by an interrupt or the like, the local processor 405
Starts and broadcast data frame 304
After creating (FIG. 20) (step 1401), the transmission path control LSI 408 is activated (step 1402).

FIG. 11 shows an example in which the number of broadcast data reception interrupts to the main processor 301 of the computer 201 is reduced by the method of this embodiment.

In this embodiment, as described above, there are a group that broadcasts at a 4 ms cycle and a group that broadcasts at a 1 ms cycle. Since both are connected to the same network, the network adapter of the computer belonging to the 4 ms cycle also receives unnecessary broadcast data of the 1 ms cycle. Figure 11
Indicates that the load on the computer belonging to the 4 ms cycle is reduced.

The transmission line control LSI 408, which has received the broadcast data from the transmission line eight times in total, receives the broadcast data from the transmission line.
Interrupt. The local processor 405 thins out the broadcast data by the logic described so far and notifies the main processor 301 of the computer 201 only in the case of the information required by the own computer. Here, 4
Only the broadcast data of the ms cycle is notified to the main processor 301, and the rest is discarded. This reduces the number of interrupts that would have entered the main processor 301 eight times to two.

[Second Embodiment] FIG. 12 shows an example in which a system is constructed as if the memory transfer is apparently performed in the same cycle between computers having different performances.

In the above-described first embodiment, an example in which different periods are provided according to performance has been described. In the second embodiment, apparently only one cycle is provided, and a computer with low performance participates in the transfer at a pace according to the performance of its own computer. The configurations of the computer and the network adapter of the second embodiment are the same as those of the first embodiment, and will be omitted. In the following, the part different from the first embodiment will be described in detail.

In FIG. 12, computer A (201A),
Computer B (201B) and Computer C (201C)
Use a high-performance computer. In addition, computer D (201D),
Computer E (201E) and Computer F (201F)
Use a low-performance computer. It is assumed that data sharing is performed by these six computers, and in this case, all of them are made to appear as if memory transfer is performed at the same cycle of 1 ms. Actually, computer D (201D), computer E (20
1E) and the computer F (201F) are systems that receive / transmit broadcast data only in a cycle of 4 ms, for example.

FIG. 13 shows the local processor 4 on the network adapter 202 when the apparent cycle is one.
It is a flowchart which shows the procedure of the reception processing of 05.

The local processor 405 is activated by a reception interrupt from the transmission path control LSI 408, and first determines whether the received data is the broadcast data 304 (step 1701). If it is not the broadcast data 304, the main processor 301 is notified as it is (step 1707). In the case of the broadcast data 304, after referring to the soft switch table 801, the memory transfer participation pace according to the performance of the own computer is confirmed (step 1702), and then the memory area tag 702 of the broadcast data 304 is analyzed (step 1703) and the counter is incremented (step 1704).

Next, it is judged whether or not the value of the counter has reached the specified number of times (step 1705). When the counter reaches the specified number, the counter is cleared (step 1706) and the main processor is notified of the reception of the broadcast data 304 (step 1707). In step 1705, if the counter value has not reached the specified number of times, the data is discarded.

In this embodiment, the soft switch table 801 stores the memory transfer participation pace according to the performance of the computer. This is, for example, the data of the specified number of times in step 1705. The data of the soft switch table 801 is set by downloading data according to the performance of the computer from the connected computer when the network adapter is initialized.

According to the present embodiment, even if the memory transfer of apparently 1 ms period is executed, it is only necessary to actually receive the thinned data once every several times. For example, 1m
In s, even if the computer does not have sufficient performance of the main processor, if broadcast data is notified to the computer once every several times, it is possible to join the network in which the memory transfer of 1 ms cycle is performed.

[Third Embodiment] FIG. 14 is an example of a system configuration in the case where computers belonging to two types of cycles exist.

Computer A (201A), Computer B (201
B) and computer C (201C) have a 1 ms period and 4
It is assumed that it has participated in the memory transfer of ms cycle. In addition, it is assumed that the computer D (201D), the computer E (201E), and the computer F (201F) participate only in the 4 ms cycle. In this embodiment, the computer A (20) that belongs to only the 1 ms period group in the first embodiment is
1A), computer B (201B), and computer C (201
C) may belong to the 4 ms period group.

Specifically, the soft switch table 801 of the network adapters connected to the computer A (201A), the computer B (201B), and the computer C (201C) describes both 1 ms and 4 ms as a cycle. I'll do it. Then, these computers belonging to both groups perform broadcast of both cycles of 1 ms and 4 ms, and the reception processing notifies the main processor of broadcast data of both cycles.

Also in this example, the computer A (201A),
The data for the 1 ms period generated by the computer B (201B) and the computer C (201C) is the computer D (201
D), computer E (201E), and computer F (201
F), the computer D (201
D), computer E (201E), and computer F (201
The broadcast data thinning in F) is effective.

[Fourth Embodiment] FIG. 15 shows an example of a system configuration in the case where segments having different periods are combined into one segment by a bridge.

In the first to third embodiments described above, it is assumed that the computers connected to one network segment are of various types (there is a difference in performance).

On the other hand, in the fourth embodiment of FIG. 15, the network (A) 203A and the network (B) 203B that originally functioned as independent networks were used.
And are cases connected by the bridge 1901. Network (A) 20 by the bridge 1901
3A and the network (B) 203B form one large network, and the broadcast data used in each network flows to another network.

Therefore, as in the first embodiment, 1
The network adapter of the computer belonging to the ms cycle segment notifies only the broadcast data of the 1 ms cycle to the main processor, and the network adapter of the computer belonging to the 4 ms cycle segment notifies only the broadcast data of the 4 ms cycle to the main processor.

Also in this example, the network (A) 20
The 1 ms cycle data of the computer A (201A), the computer B (201B), and the computer C (201C) that have flowed from the 3A to the network (B) 203B are stored in the computer D (20).
1D), computer E (201E), and computer F (201
F) can be thinned out, and the processing load for receiving broadcast data can be reduced.

[Fifth Embodiment] FIG. 16 is a system configuration example in the case where there is a computer that does not participate in memory transfer in the network.

Computer A (201A), Computer B (201
B) and computer C (201C) join the memory transfer of 1 ms period, and computer E (201E) and computer F
(201F) participates in a 4 ms cycle. Calculator D
(201D) does not participate in memory transfer.

Since the computer D (201D) has not joined the memory transfer, the memory switch cycle is not described in the soft switch table 801 of the network adapter of the computer D (201D). Therefore, if processing is performed as shown in FIG. 1 as in the case of the first embodiment, all broadcast data for memory transfer are discarded and the main processor of the computer D (201D) is not notified.

Also in this example, the computer A (201A),
Data for the 1 ms period generated by the computer B (201B) and the computer C (201C), and the computer E (201
E) and the data for the 4 ms period generated by the computer F (201F) arrives at the computer D (201D).
The discard of broadcast data in the computer D (201D) according to the present invention is effective.

[0085]

The network adapter and the network system according to the present invention have the following effects.

By thinning out the broadcast data for matching the contents of the transfer memory at the network adapter level, it is possible to reduce the number of broadcast data reception interrupts to the main processor of the computer and reduce the matching processing load of the main processor. it can.

Further, as a concrete method thereof, the local processor on the adapter can judge whether or not the data should be thinned out by reading the memory area tag in the broadcast data upon reception.

As another specific method, the local processor on the adapter notifies the main processor of data once every n times by counting the number of times of reception of broadcast data, and the subsequent data is Can be thinned out.

Further, in the computer adapter not participating in the memory transfer, the communication processing load of the main processor can be reduced by discarding all broadcast data and not notifying the main processor.

Further, by making it possible to set the frequency of thinning out the broadcast data with a switch, a common adapter can be used for a plurality of types of computers.

Further, as a concrete method thereof, by providing a soft switch table which can be downloaded from the computer side, the local processor can judge whether or not to interrupt the main processor by referring to the table.

Due to the above concrete effects, the matching processing load of the cycle transfer memory in the computer main processor can be reduced, and a computer with a low performance can be used as a cycle among computers having different main processor performances such as a minicomputer, a personal computer, and WS. The memory transfer can be executed without adjusting.

[Brief description of drawings]

FIG. 1 is a flowchart of local processor reception processing on a network adapter according to a first embodiment of this invention.

FIG. 2 is a network system configuration example of the first embodiment.

FIG. 3 is a conceptual explanatory diagram of a periodic memory transfer.

FIG. 4 is a hardware configuration diagram for realizing the first embodiment.

FIG. 5 is a diagram showing sharing of memory transfer processing in the first embodiment.

FIG. 6 is a memory transfer realization diagram of the first embodiment.

FIG. 7 is a system configuration diagram in which two cycles are provided according to the performance of the main processor.

FIG. 8 is a flowchart of a computer main processor reception process according to the first embodiment.

FIG. 9 is a flowchart of a computer main processor transmission process according to the first embodiment.

FIG. 10 is a flowchart of a local processor transmission process on the network adapter according to the first embodiment.

FIG. 11 is an explanatory diagram of reducing the number of interrupts according to the method of the first embodiment.

[Figure 12] System configuration diagram with a single apparent cycle

FIG. 13 is a flowchart of local processor reception processing on the network adapter when the number of apparent cycles is one.

FIG. 14 is a system configuration diagram when there are computers belonging to two cycles.

FIG. 15 is a system configuration diagram in the case where segments having different periods are combined into one segment by a bridge.

FIG. 16 is a system configuration diagram when there is a computer that does not participate in memory transfer.

FIG. 17 is a hardware configuration diagram of conventional memory transfer.

FIG. 18 is a diagram showing a communication processing share of conventional memory transfer.

FIG. 19: Conventional memory transfer system diagram

FIG. 20 is a diagram showing a broadcast data (BD) frame format for memory transfer information.

[Explanation of symbols]

201 computer 202 network adapter 203 network 301 computer main processor 302 memory transfer area 303 computer # 1
Memory transfer area 304 Broadcast data 401 Computer main memory 402 Computer system bus 403 Conventional network adapter 404 Local memory 405 Local processor 406 Transmit / receive buffer 407 Local bus 408 Transit control LSI 701 Memory transfer data 702 Memory area tag 703 Memory transfer information 704 MAC header 801 Soft switch table 1901 Bridge

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Kenji Kataoka 1099, Ozenji, Aso-ku, Kawasaki-shi, Kanagawa, Ltd. System Development Laboratory, Hitachi, Ltd. (72) Inventor, Osamu Takada 1099, Ozen-ji, Aso-ku, Kawasaki, Kanagawa Hitachi, Ltd. System Development Laboratory (72) Inventor Hiroshi Wataya 5-2-1 Omika-cho, Hitachi-shi, Ibaraki Hitachi Ltd. Omika factory

Claims (12)

[Claims] 1. A plurality of computers each having a transfer memory area for sharing data are connected to a network,
Each computer periodically broadcasts the data of its own computer area in the transfer memory area to other computers, and the computer receiving the broadcast data stores the data in the corresponding computer area of the transfer memory area. As a result, in a system for sharing data by matching the contents of the transfer memory area among computers, a network adapter for connecting each computer to a network, and thinning out the received periodic broadcast data. A network adapter comprising means. 2. The thinning means discriminates a broadcast cycle of received broadcast data, and when the cycle is a cycle of broadcast data to be taken in by the computer, the received data is notified to a processor of the computer. However, if not, the received data is discarded, and the network adapter according to claim 1. 3. The thinning means counts the number of times of reception of broadcast data, notifies the received data to a processor of its own computer only when a predetermined number of times is reached, and discards other received data. The network adapter according to claim 1, wherein the network adapter is a network adapter. 4. The thinning means executes matching of the contents of the transfer memory area in a plurality of other computers, and when the own computer does not participate in the matching of the transfer memory, the received transfer memory match. The network adapter according to claim 1, wherein all the broadcast data for conversion is discarded and the processor of the own computer is not notified. 5. The thinning means comprises a switch for setting information for determining whether to notify the received broadcast data to the processor of the own computer or to discard it without notifying the processor. 1. The network adapter described in 1. 6. The network adapter according to claim 5, wherein the switch is a hardware switch. 7. The network adapter according to claim 5, wherein the switch is a soft switch table downloadable from the computer side to the network adapter. 8. A network system comprising a plurality of computers, a network adapter connected to each of the plurality of computers, and a network to which the network adapter is connected, wherein the plurality of computers share data. , Each of the computers has a transfer memory area for storing data shared with another computer, and the computer in the transfer memory area periodically broadcasts the data of the area for the own computer to the other computers. A means for activating the connected network adapter, and means for storing broadcast data from another computer received via the network adapter connected to the own computer in the other computer area of the transfer memory area , The network adapter, the periodic broadcast in response to the startup And a method of executing the broadcast request and the broadcast data from another computer is received, it is determined whether or not the broadcast data should be captured in the computer, and the broadcast data to be captured is notified to the computer and the broadcast data that is not likely is discarded. Accordingly, a network system comprising means for thinning out received broadcast data. 9. The thinning means counts the number of times of receiving broadcast data, notifies the received data to the processor of its own computer only when the number of times reaches a predetermined number, and discards other received data. The network system according to claim 9, wherein: 10. The plurality of computers are composed of various computers having different main processor performances, and the thinning means of the network adapter determines a thinning rate based on the performance of the computer connected to the network adapter. The network system according to claim 8. 11. A plurality of computers, a network adapter connected to each of the plurality of computers, and a network to which the network adapter is connected, wherein the plurality of computers share a plurality of data. In a network system in which a plurality of groups each including computers are defined to allow duplication of computers, each of the computers includes a transfer memory area for storing data shared with another computer, and a transfer memory area of the transfer memory area. In order to periodically broadcast the data for the local computer area of the local computer to other computers at a cycle unique to the group to which the local computer belongs, a means for starting the network adapter connected to the local computer and the means connected to the local computer. Broadcast data received from another computer via the network adapter And a means for storing the storage area in the other computer area of the memory area, the network adapter executing the periodic broadcast according to the start-up, and receiving broadcast data from another computer. , The broadcast data is received by discriminating the cycle of the broadcast data, notifying the broadcast data to the computer when the cycle is a cycle unique to the group to which the computer belongs, and discarding the broadcast data other than that. A network system comprising means for thinning out data. 12. The computer according to claim 11, wherein the plurality of computers are composed of various computers having different main processor performances, and in the group, computers having similar performances are grouped together. Network system.