CN1963761A - Dummy multiprocessor system based on network and processing method - Google Patents
Dummy multiprocessor system based on network and processing method Download PDFInfo
- Publication number
- CN1963761A CN1963761A CN 200510048517 CN200510048517A CN1963761A CN 1963761 A CN1963761 A CN 1963761A CN 200510048517 CN200510048517 CN 200510048517 CN 200510048517 A CN200510048517 A CN 200510048517A CN 1963761 A CN1963761 A CN 1963761A
- Authority
- CN
- China
- Prior art keywords
- packet
- instruction
- main control
- computer
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Computer And Data Communications (AREA)
Abstract
This invention relates to virtual multiple processor system and its method based on network, wherein, the system comprises only one large user task organization composed of programs of multiple processor system based on internet and to align program orders to other customer machine to process main control host; at least one machine to receive orders from main control orders then going back process results or to compute middle status customer machine; the main control machine and customer machine are connected by computer. This invention method comprises the following steps: main control host aligns task to network customer processor and the processor sends the results to computer status to main control machine.
Description
Technical field:
The present invention relates to computer science, technology such as it and multicomputer system, computer network, operating system are closely related, be a kind of that a plurality of computer systems in the network are virtual for multicomputer system, Task Distribution, the task that multicomputer system must finish handled to finish, the technology of virtual processor scheduling.
Background technology:
Multicomputer system is a kind of Computer Systems Organization form of parallel processing calculation task, uses the computer system of this structure construction to cost an arm and a leg, and can not be accepted by individual or general unit.In addition, the processor of multicomputer system always connects by a kind of topological structure, and its manageable task is often closely related with the sort processor topological structure.
Along with popularizing and development of computer networking technology, communication speed is more and more faster between the computer system that interconnects in the network, and the price of single computer systems is but more and more lower, and performance is more and more higher.Personal computer system's development makes individual or unit only use personal computer system's one small part resource to meet the demands, and resource for computer system exists the phenomenon of significant wastage.And existing network technology generally can be counted as a kind of Internet resources shared system, and network service just realizes a kind of means of resource sharing, and the potential concurrency that exists in the network can not be effectively played.The quickening of network service speed is in conjunction with the idle existence of a large amount of resources for computer system in the network, but determined a modern computer network system to exist the ability of potential parallel processing mass computing task.
Summary of the invention:
Utilized fully for making the active computer resources in network, and on handling, large-scale task brings into play its potential parallel advantage, the invention provides a kind of based on network dummy multiprocessor system and disposal route, utilize the high-speed communication of existing network, main control computer, promptly is distributed on the client computer and handles Task Distribution other computer system in the network by network; Compare with existing single processor system, the present invention will improve large-scale task handling speed greatly; Compare with existing physics multicomputer system, cost of the present invention is much cheap again.
For achieving the above object, the present invention has adopted following technical scheme:
Described based on network dummy multiprocessor system, it comprises: one and only one large-scale user task is organized as the program is made up of the instruction of based on network dummy multiprocessor system, also the command assignment of this program removed the main control computer handled on other client computer; At least one is used to receive from the instruction of main control computer and handles, returns afterwards result or calculate the client computer that intermediateness is given main control computer; Main control computer is connected with client's machine computer network.
Described based on network dummy multiprocessor system, its main control computer comprises: the program storing unit that is used to the program of depositing; Be used for client computer distributor to client computer round-robin queue access client information; Be used for the instruction client computer packing device that instructs and client information is packed and encapsulated; Be used for sending the packet transmitter of packet to client computer; Be used to receive data packet receiver from the client data bag; Be used for de-packetizer that the packet that receives is analyzed; Wherein, instruction client computer packing device takes out a client information that is in state to be allocated from the client computer distributor, get an instruction from the program storing unit, and the two is packaged into packet, passes to the packet transmitter and sends to corresponding client computer; After data packet receiver receives packet from client computer, this packet is sent into de-packetizer, and de-packetizer obtains the result of client information and instruction after to this data packet analysis or calculates intermediateness: client information is sent into client computer round-robin queue in the client computer distributor again; The intermediateness of instruction process result or calculating is sent into the program storing unit, corresponding instruction or data variable group are made amendment.
Described based on network dummy multiprocessor system, its client computer comprises: be used to receive the data packet receiver from the main control computer packet; Be used to analyze the reception de-packetizer of received data packet; Be used to execute instruction and bear results or calculate the instruction execution unit of intermediateness; Be used for to this machine information with from the result of instruction execution unit or calculate the reception packing device that intermediateness is packaged into the packet that can be received by main control computer; Be used for packaged packet is sent to the packet transmitter of main control computer; Wherein, after data packet receiver is received the packet that sends from main control computer, give the reception de-packetizer with this data packet delivery, by receiving de-packetizer it is analyzed, therefrom obtain instruction, instruction execution unit is sent in instruction carried out, carry out the back by instruction execution unit result or calculating intermediateness are sent into the reception packing device, receive packing device with this result or calculate intermediateness and be packaged into packet with this machine information, and with this data packet delivery to the packet transmitter, send to main control computer.
A kind of disposal route of based on network dummy multiprocessor, the method step that its main control computer and client computer connect comprises:
1), after the main control computer start, soon this machine information and send into instruction client computer packing device and be packaged into packet, passes to packet transmitter broadcast transmission to network;
2) after, the data packet receiver of other computer system that exists on the network is received this packet, this packet is sent into the reception de-packetizer, receiving de-packetizer analyzes this packet and obtain main control computer information in the packet: the main control computer information that obtains is analyzed in this computer system utilization, connect with main control computer, and this machine information sent into receive packing device and be packaged into the connection request packet, pass to the packet transmitter, send to main control computer by the packet transmitter;
3), the data packet receiver of main control computer is received behind the connection request packet that other computer system is sent, this bag is sent into de-packetizer analysis, the computerized information that de-packetizer obtains analysis is sent into the client computer round-robin queue in the client computer distributor, and the represented computer system of computerized information that enters client computer round-robin queue promptly becomes a client computer of main control computer.
The disposal route of described based on network dummy multiprocessor, the storage means of its client information comprises the following steps:
(1), after main control computer receives the packet of other computer system on the network, this packet is sent into de-packetizer, the computerized information that de-packetizer obtains analysis is sent into the client computer round-robin queue of client computer distributor;
(2), in client computer round-robin queue, described computerized information is received the precedence composition formation of packet that computing machine sends according to main control computer, on article one information of this formation head pointer is set, go out team with the control client information, on the last item information of this formation, tail pointer is set, joins the team with the control client information; The team's action of joining the team out is undertaken by group direction of joining the team out: after the information of head pointer indication was removed, head pointer moved a position backward, points on next bar information position; When a fresh information is sent into, this fresh information is queued at the end of queue, tail pointer is moved a position backward, point on this fresh information position.
The disposal route of described based on network dummy multiprocessor after main control computer and client computer connect, to the process method step of user task is:
Instruction client computer packing device in a, the main control computer, from the program storing unit, take out an instruction, from the client computer distributor, take out a client information that is in state to be allocated, the two is packaged into packet, send into the packet transmitter and send to client computer;
After the data packet receiver of b, client computer is received packet, it is passed to the reception de-packetizer parse instruction, sending into instruction execution unit handles, instruction execution unit is sent the result of instruction or the intermediateness of calculating into the reception packing device, packing with this client information is packaged into packet, sends into the packet transmitter and sends to main control computer;
After c, described main control computer receive packet from client computer, the packet that receives is sent into de-packetizer, after de-packetizer is analyzed, handle respectively according to the information that data packet analysis is obtained: de-packetizer obtains client information and instruction execution result after to data packet analysis or calculates intermediateness, client information is sent into again the client computer round-robin queue of client computer distributor, instruction execution result or calculating intermediateness are sent in the program storing unit, and to wherein command adapted thereto or data variable group are made amendment;
D, enter step a, carry out next step command assignment.
Described based on network dummy multiprocessor disposal route, the wherein instruction of from described program storing unit, taking out, this instruction is made up of other director data positional information that instruction title, data, instruction results may influence, and this instruction also comprises command status information; The instruction of taking out from described program storing unit, this instruction also can be made up of instruction title, data variable group.
In sum, the present invention utilizes network system of the prior art just, provides a kind of by network technology dummy multiprocessor system tasks in parallel treatment technology and disposal route that realize, that make full use of the idle computer resource that exists in the network system.In the present invention, network service is system control information and the transfer mode of data in network, network service makes that the distributed computers system becomes as a whole in the network, and they are coordinated with each other, resembles the large-scale user task of parallel processing the multicomputer system.They can not only simulate the data processing method of multicomputer system, and do not have the restriction of the processor connection topological structure of real multicomputer system, and system resource has also obtained utilizing fully, and the real multicomputer system of system cost is low.Data-searching shows, does not also find the research of this respect.
Utilize technique scheme, a kind of based on network dummy multiprocessor system of the present invention and disposal route are not only newly split a technology outlet for big task handling, more make the cost of total system be based upon a kind of can be on the hardware platform that people accepted.In sum, realize the present invention, to reach following good effect: can not only remedy the expensive deficiency of multicomputer system, more brought into play the advantage of network distribution type parallel computation, make full use of the computer resource that computer network exists, bring into play their effect, the parallel task processing that is reflected in the present invention has bigger degree of freedom, and almost being defined on present all parallel processor systems of task can both obtain handling in system involved in the present invention.
Description of drawings:
Figure 1 shows that the network connection diagram of main control computer and client computer;
Figure 2 shows that information transfering relation synoptic diagram between main control computer and the client computer;
Figure 3 shows that the workflow diagram of main control computer;
Figure 4 shows that the workflow diagram of client computer;
Figure 5 shows that based on network dummy multiprocessor system implementation structure synoptic diagram;
Figure 6 shows that client computer round-robin queue management synoptic diagram;
Figure 7 shows that main control computer processing procedure embodiment schematic flow sheet;
Figure 8 shows that client processes process embodiment schematic flow sheet.
Embodiment:
A kind of based on network dummy multiprocessor system provided by the invention and disposal route provide a kind of by network technology dummy multiprocessor system tasks in parallel treatment technology and disposal route that realize, that make full use of the idle computer resource that exists in the network system.
This method at first with a computer system in the network as main control computer (1), to decompose large-scale user task is the basic task that based on network dummy multiprocessor system can be handled, and they are distributed to the processing of getting on of other computer system except that main control computer (1), to obtain bigger system effectiveness.Other computer system described here is client computer, and client computer (2) is to handle the computer system of the basic task that is sent by main control computer (1), and they receive from the control information and the data of main control computer (1) and handle.
Figure 1 shows that the network connection diagram of main control computer of the present invention and client computer, the hardware platform that the present invention relied on is to have an operating main control computer (1) and many operating client computer (2) at least in a computer network that connects in any way and this network.Figure 2 shows that information transfering relation synoptic diagram between main control computer and the client computer, reflect the information transfering relation between main control computer of the present invention (1) and the client computer (2): main control computer (1) sends control information, computations or data to client computer (2), and client computer (2) is returned to main control computer (1) and calculated intermediateness or result of calculation.
Just because of there is annexation shown in Figure 1, information transmission between main control computer (1) and the client computer (2) could realize, again because the computer system among Fig. 1 is implemented the specifying information transmission by the requirement of Fig. 2, a based on network dummy multiprocessor system just is achieved simultaneously.
As shown in Figure 3, the workflow of main control computer (1) is: main control computer (1) is broadcasted this machine information on network, simultaneously by monitoring network, find to exist in the network operating and beam back the computer system of link information, they are registered, and the system after the registration is as the client computer (2) of based on network dummy multiprocessor system; In system, there is client computer (2), during that exist to need handle in the simultaneity factor, after decomposing, the user task that can on based on network dummy multiprocessor system, handle, these task distribution got on to finish to client computer (2); No matter be when above-mentioned work disposal or after handling, main control computer (1) all will continue to monitor network, to find client computer.
Client computer (2) is that the computer system of moving from a network changes.The computer system that can change client computer (2) into, when it receives main control computer (1) information, utilize main control computer (1) information, send link information to main control computer (1), behind the response message that receives from main control computer (1), promptly transfer to and monitor network, carry out this calculation task and return execution result or the calculating intermediateness to find calculation task from main control computer (1).When a computer system receives response message in the network, illustrate that it is found, registers and be converted into a client computer (2) by main control computer (1).After sending link information, then need to continue to send out link information to main control computer (1) as if the response message that can not receive in time from main control computer (1) to main control computer (1), so that possible afterwards can be received and be converted into client computer (2) by main control computer (1) sometime, participate in based on network dummy multiprocessor system.
The task that can be assigned to the last operation of client computer (2) is that some design at based on network dummy multiprocessor system, and than the little basic task of user task, they can be described by the instruction that table one and table three provide.
The instruction relevant with the dummy multiprocessor system that the present invention is based on network has two kinds: the based on network dummy multiprocessor system of being represented by other director data positional information of instructing title, data, instruction results to influence and command status shown in the table one instructs, instruct with the based on network dummy multiprocessor system of representing with the data variable group shown in the table three, they all can be realized and cut both ways.
Instruction as shown in Table 1 is made up of other director data positional information and command status that instruction title, data, instruction results may influence; Wherein, command status comprises ready, wait two states, and the former presentation directives's data are complete, and the latter represents incomplete.When program adopts this instruction, instruction will be stored in the program storing unit (3) item by item, and when instruction fetch, whole piece instructs and read item by item.
Instruction as shown in Table 3 is made up of instruction title and data variable group, when adopting the program of this instruction to deposit in the program storing unit (3), being divided into two parts deposits, part store data set of variables, a part is deposited instruction one by one, comprises name variable in the data variable group in the instruction.When instruction fetch from program storing unit (3), produce and read instruction and two actions of read data set of variables.When the program of this instruction formation is read and write, processing procedure is than the instruction complexity shown in the processing list one, but it is few to deposit the program of forming than the instruction shown in the table one in the shared space of this program, and existing in addition single processor system program is similar to this program format.
The program that instruction constituted of these two kinds of forms can both be accomplished in technical scheme of the present invention.For ease of explanation, below the program that instruction constituted shown in the employing table one is described among the embodiment.
The instruction of based on network dummy multiprocessor system is a kind of basic functional units that can carry out on based on network dummy multiprocessor system, and it is could be on based on network dummy multiprocessor system processed or carry out that large-scale user task must be broken down into the instruction of based on network dummy multiprocessor system.The instruction class of based on network dummy multiprocessor system is similar to the computer instruction on the single processor system, they constitute the user program that can on based on network dummy multiprocessor system, move.
A based on network dummy multiprocessor system program is by decomposing, finally constitute by the instruction of based on network dummy multiprocessor system, and the instruction of based on network dummy multiprocessor system by make up too can primordial in the dummy multiprocessor system program of network.For convenience of explanation, instruction shown in the table one is reduced to the form that table two provides, based on network dummy multiprocessor system instruction after the simplification shown in the table two is made of command content and command status, and one of command content has just comprised other director data positional information that the instruction title shown in the table one, data, instruction results may influence.
Notice that the based on network dummy multiprocessor system instruction shown in the table one structurally is essentially different with the traditional computer instruction.As, the i386 instruction
ADD?AX，BX
ADD is the instruction title, a kind of function that can realize in the corresponding hardware, AX and BX deposit the data of participation calculating and the register of result of calculation, the data in the based on network dummy multiprocessor system instruction shown in their difference corresponding tables one, but different with it.ADD AX, the meaning of BX is delivered among the AX for the result with AX+BX, if ADD is AX, AX is used in the later instruction of BX, then instructs ADD AX, the result of calculation of BX just to have influence on this instruction, AX plays the effect of transmitting this influence.And the based on network dummy multiprocessor system instruction shown in the table one of the present invention is not based upon on one group of registers group that defines, and has just provided Data Position in instruction.In the based on network dummy multiprocessor system instruction shown in the table one, what corresponding Data Position was deposited is exactly the needed data of this instruction manipulation, after instruction is finished, its result is also placed on the relevant Data Position, but need indicate this result and will influence the data on which Data Position of which bar instruction of back.
| The instruction title | Data | Other director data position that instruction results may influence | Command status |
The based on network dummy multiprocessor system instruction that table one is represented by other director data position of instructing title, data, instruction results to influence and command status
| Command content | Command status |
Based on network dummy multiprocessor system instruction after instruction shown in table two table one is simplified
| The instruction title | The data variable group |
The based on network dummy multiprocessor system instruction that table three is represented with the data variable group
Certainly, with the corresponding based on network dummy multiprocessor system instruction of conventional single-processor computers system directive, as shown in Table 3, comprise instruction title and data variable group.Variable in the data variable group is equivalent to the register in the traditional computer instruction, and they are Data transmission between the instruction of based on network dummy multiprocessor system program.Obviously, the instruction of instruction shown in the table three and conventional computer system has natural sibship, and they are closely similar, but different.At first the variable in the data variable group does not have good hardware corresponding with it, and they are the same with the variable effect that defines in the C language, are a kind of necessary setting of based on network dummy multiprocessor system on the realization task is handled.But adopt the instruction shown in the table three, variable initial value is under 0 the situation, can economize the content of command status in directive command of sketch form in the setting data set of variables.Reason is very simple, the variable in all data variable groups, and its value just is ready to when initialization, is 0, and their value has just been revised in based on network dummy multiprocessor system program implementation.Simultaneously and since utilize the data variable group can be in based on network dummy multiprocessor system program Data transmission, thereby " other director data positional information that instruction results may influence " in need not table one directive command.
Client computer (2) is based on the hardware carrier of the dummy multiprocessor system instruction execution of network, they can discern all based on network dummy multiprocessor system instructions, press command request and carry out, and by the requirement of instructing, instruction execution result is write back return main control computer (1) in the instruction.The implementation of client computer (2) as shown in Figure 4, client computer is sent out link information to main control computer, when receiving the main control computer response message, monitors network, finds and carries out from the task of main control computer and return execution result; When not receiving the main control computer response message, continue to send out link information to main control computer.Client computer specific implementation method is as follows:
Client computer (2) is by the packet of data packet receiver (7) reception from main control computer (1), and it is passed to receive de-packetizer (9), from packet, decomposite instruction and be sent to instruction execution unit (11) by receiving de-packetizer (9), after instruction execution unit (11) is carried out, instruction execution result or calculating intermediateness are sent into reception packing device (10), receive packing device (10) with instruction execution result or calculate intermediateness and this machine information is packaged into packet, pass to packet transmitter (8), this bag is sent to main control computer (1); Wherein, receiving packing device (10) can be with two kinds of packing data: 1. this client computer machine information; 2. the execution result that send over after will instructing and carrying out of this client computer machine information and instruction execution unit (11) or calculate intermediateness.If have only this client information in the packet, after then the de-packetizer (5) of main control computer (1) solves this information, this information is sent into client computer round-robin queue in the client computer distributor (14), (generally speaking, only comprise this machine information in the packet that other computer system and main control computer (1) are sent out when connecting in the network, promptly become the client computer of main control computer (1) after computer system and main control computer (1) connect); If the information in the packet had both included client information, include result or intermediateness information that instruction execution unit (11) sends over again, then the de-packetizer (5) of main control computer (1) not only will send to this client information client computer distributor (14), also to or calculate intermediateness and send into program storing unit (3) simultaneously, revise dependent instruction or related data variable instruction execution result.
Based on network dummy multiprocessor system implementation structure synoptic diagram of the present invention as shown in Figure 5.
The defined packet of the present invention sees Table four, the character of the concrete data of data type specification, such as data type with non-0 natural number coding, the concrete data of 0 expression are this machine information of main control computer (1), the concrete data of 1 expression are this machine information of client computer, the concrete data of 2 expressions are instruction, and the concrete data of 4 expressions are the client computes result, and the concrete data of 5 expressions are client computes intermediateness etc.
| Data type | Concrete data |
Table four packet
Main control computer (1) is being undertaken user task is being organized as the program that based on network dummy multiprocessor system can be handled, and control program carries out on based on network dummy multiprocessor system, therefore, its design more complicated, specific as follows.
The workflow of main control computer (1) as shown in Figure 3, main control computer is monitored network, finds client computer and registration, judges the user task that whether exists in the main control computer after the decomposition then, if exist, then allocating task is gone processing to client computer; If do not exist, then continue to monitor network, find client computer and registration.
Should there be array apparatus down in addition, to finish the required function of finishing of main control computer (1):
1. program storing unit (3): based on network dummy multiprocessor system program should leave in the main control computer (1) with the receptible form of main control computer (1).The based on network dummy multiprocessor system instruction of composition program can be formed (as shown in Table 1) by other director data positional information and command status that instruction title, data, instruction results may influence, can also form (as shown in Table 3) by instruction title, data variable group.
2. instruction fetch and send to the implement device of client computer (2) is called for short to get and refers to and dispensing device: gets and refer to and dispensing device includes instruction client computer packing device (4), client computer distributor (14), packet transmitter (12).The at first instruction fetch from program storing unit (3) of instruction client computer packing device (4), from client computer distributor (14), get a current client information that is in state to be allocated, then the two is packaged into packet, passes to packet transmitter (12) and send to client computer (2).
3. receive the return data device
Receive the return data device and comprise data packet receiver (13), de-packetizer (5), implement in two kinds of situation: when other director data positional information and the command status that may be influenced by instruction title, data, instruction results shown in the based on network dummy multiprocessor system employing table one are formed the situation of instructing; Or when the situation shown in the based on network dummy multiprocessor system employing table three by the instruction of instructing title, data variable group to form; No matter be any situation, after data packet receiver (13) is received packet from client computer (2), give de-packetizer (5) with this data packet delivery, de-packetizer (5) is analyzed this packet, the client information that wherein comprises is branched away, send into the client computer round-robin queue in the client computer distributor (14); For instruction execution result or calculating intermediateness data, de-packetizer (5) is at different situations, and the value of the middle variable of the dependent instruction in the update routine storing unit (3), or modification data variable group (6) is respectively correctly carried out with the assurance program.
4. client computer distributor
Storage is by the round-robin queue that is made up of client information in the client computer distributor (14), queue size is by the decision of the network size of based on network dummy multiprocessor system, all unit between from the head pointer to the tail pointer all depositing can allocating task client information.When instructing client computer packing device (4) when reading client information, head pointer goes out team along group direction of joining the team out with a client information, gives instruction client computer packing device (4); When de-packetizer (5) passes over the client information that analyzes, this client information is sent into rear of queue, and mobile tail pointer points to the information of newly sending into.
Figure 6 shows that client computer round-robin queue management synoptic diagram, wherein, a client computer is all represented in each unit in the sector region.
The concrete condition that client information is sent in the client computer round-robin queue is, described computerized information is received the precedence composition formation of packet that computing machine sends according to main control computer, on article one information of this formation head pointer is set, go out team with the control client information, on the last item information of this formation, tail pointer is set, joins the team with the control client information; The team's action of joining the team out is undertaken by group direction of joining the team out: after the information of head pointer indication was removed, head pointer moved a position backward, points on next bar information position; When a New Client information is sent into, this fresh information is queued at the end of queue, tail pointer is moved a position backward, point on this New Client information position.
Above-mentioned four parts are formed the device for carrying out said of main control computer (1).
| Client information | Command information |
Table five main control computer sends packet
| Client information | Result data |
Table six client computer sends packet
In addition, table five is depicted as the packet that main control computer (1) sends to client computer (2), table six is depicted as the packet that client computer (2) sends to main control computer (1), packet shown in table five and the table six is after main control computer (1) connects with client computer (2), and main control computer (1) sends to packet (table six) main control computer (1), that comprise this client information and instruction execution result after sending to packet (table five) client computer (2), that comprise client information and instruction and client computer (2) execution command.And when main control computer (1) and client computer (2) when connecting, main control computer (1) broadcast transmission only comprises main control computer (1) information to the packet on the network; After client computer (2) is received the broadcast data packet of main control computer (1), utilize this main control computer (1) information, will send the connection request packet to main control computer (1) after connecting with main control computer (1), only comprise the information of client computer (2) in this connection request packet.
In sum, a kind of based on network dummy multiprocessor system provided by the present invention comprises: one and a main control computer (1) only, be used for large-scale user task be organized as the program formed by based on network dummy multiprocessor system instruction, also with the command assignment of this program to other client computer (2) processing of getting on; At least one client computer (2) is used for receiving from the instruction of main control computer (1) and handles, and returns result afterwards or calculates intermediateness and give main control computer (1); Main control computer (1) is connected with computer network with client computer (2).
Described main control computer (1) comprising: the program storing unit (3) that is used to the program of depositing; Be used for client computer distributor (14) to client computer round-robin queue access client information; Be used for the instruction client computer packing device (4) that instructs and client information is packed and encapsulated; Be used for sending the packet transmitter (12) of packet to client computer (2); Be used for receiving data packet receiver (13) from client computer (2) packet; Be used for de-packetizer (5) that the packet that receives is analyzed; Wherein, instruction client computer packing device (4) takes out a client information that is in state to be allocated from client computer distributor (14), from program storing unit (3), get an instruction, the two is packaged into packet, passes to packet transmitter (12) and send to corresponding client computer (2); After data packet receiver (13) receives packet from client computer (2), this packet is sent into de-packetizer (5), and de-packetizer (5) obtains the result of client information and instruction after to this data packet analysis or calculates intermediateness: client information is sent into client computer round-robin queue in the client computer distributor (14) again; The intermediateness of instruction process result or calculating is sent into program storing unit (3), corresponding instruction or data variable group are made amendment.
Described client computer (2) comprising: be used for receiving the data packet receiver (7) from main control computer (1) packet; Be used to analyze the reception de-packetizer (9) of received data packet; Be used to execute instruction and bear results or calculate the instruction execution unit (11) of intermediateness; Be used for to this machine information with from the result of instruction execution unit (11) or calculate the reception packing device (10) that intermediateness is packaged into the packet that can be received by main control computer (1); Be used for packaged packet is sent to the packet transmitter (8) of main control computer (1); Wherein, after data packet receiver (7) is received the packet that sends from main control computer (1), give reception de-packetizer (9) with this data packet delivery, by receiving de-packetizer (9) it is analyzed, therefrom obtain instruction, instruction execution unit (11) is sent in instruction to be carried out, carry out the back by instruction execution unit (11) result or calculating intermediateness are sent into reception packing device (10), receive packing device (10) with this result or calculate intermediateness and this machine information is packaged into packet, and give packet transmitter (8) with this data packet delivery, send to main control computer (1).
Give client computer (2) before in main control computer (1) allocating task, other computer system that needs to exist in elder generation and the network connects, make the computer system that connects become the client computer (2) of main control computer (1), provided the method step that main control computer and client computer connect below:
1), after main control computer (1) start, soon this machine information and send into instruction client computer packing device (4) and be packaged into packet, passes to packet transmitter (12) broadcast transmission to network;
2) after, the data packet receiver of other computer system that exists on the network (7) is received this packet, this packet is sent into reception de-packetizer (9), receiving de-packetizer (9) analyzes this packet and obtains main control computer information in the packet, this computer system utilization is analyzed the main control computer information and the main control computer (1) that obtain and is connected, and this machine information sent into receive packing device (10) and be packaged into the connection request packet, pass to packet transmitter (8), send to main control computer (1) by packet transmitter (8);
3), the data packet receiver (13) of main control computer (1) is received behind the connection request packet that other computer system is sent, this bag is sent into de-packetizer (5) to be analyzed, the computerized information that de-packetizer (5) obtains analysis is sent into the client computer round-robin queue in the client computer distributor (14), and the represented computer system of computerized information that enters client computer round-robin queue promptly becomes a client computer (2) of main control computer (1).
After main control computer (1) connected with client computer (2), main control computer (1) need be stored and manage client computer (2) information that connects in the tabulation, has provided the storage means of a kind of client computer (2) information below, and step is as follows:
(1), after main control computer (1) receives the packet of other computer system on the network, this packet is sent into de-packetizer (5), described other computerized information that de-packetizer (5) obtains analysis is sent into the client computer round-robin queue in the client computer distributor (14);
(2), in client computer round-robin queue, described other computerized information is received the precedence composition formation of the packet of described other computing machine according to main control computer, on article one information of this formation head pointer is set, go out team with the control client information, on the last item information of this formation, tail pointer is set, joins the team with the control client information; The team's action of joining the team out is undertaken by group direction of joining the team out: after the information of head pointer indication was removed, head pointer moved a position backward, points on next bar information position; When a fresh information is sent into, this fresh information is queued at the end of queue, tail pointer is moved a position backward, point on this fresh information position.
After main control computer (1) and client computer (2) connect, be to the process method step of user task:
Instruction client computer packing device (4) in a, the main control computer (1), from program storing unit (3), take out an instruction, from client computer distributor (14), take out a client information that is in state to be allocated, the two is packaged into packet, sends into packet transmitter (12) and send to client computer (2);
After the data packet receiver (7) of b, client computer (2) is received packet, it is passed to reception de-packetizer (9) parse instruction, sending into instruction execution unit (11) handles, instruction execution unit (11) is sent the result of instruction or the intermediateness of calculating into reception packing device (10), packing with the information of this client computer is packaged into packet, sends into packet transmitter (8) and sends to main control computer (1);
After c, described main control computer (1) receive packet from client computer (2), the packet that receives is sent into de-packetizer (5) analysis, according to de-packetizer (5) information that data packet analysis obtains is handled respectively: de-packetizer (5) is to obtaining client information and instruction execution result or calculating intermediateness behind the data packet analysis, client information is sent into again the client computer round-robin queue of client computer distributor (14), instruction execution result or calculating intermediateness are sent in the program storing unit (3), and to wherein command adapted thereto or data variable group are made amendment;
D, enter step a, carry out next step command assignment.
Described based on network dummy multiprocessor disposal route, the wherein instruction of from described program storing unit (3), taking out, this instruction is made up of other director data positional information and command status information that instruction title, data, instruction results may influence; The instruction of taking out from described program storing unit (3), this instruction also can be made up of instruction title, data variable group.
For verifying realizability of the present invention, the inventor utilizes the Socket programming successfully to realize the experimental system of the present invention by Fig. 5 design on Linux.In this experimental system:
1. network service utilizes the socket programming to realize.
2. the packet definitions of transmitting information between main control computer (1) and the client computer (2) is as follows:
struct?package
{ ing fnc_class; Data type
Struct sockaddr_in ip; This machine information: adopt the IP address in this example
Struct ID_opc; Instruction
{
Char opc[128]; The instruction title
Struct data; Director data
{
Char a; Command status
Double b; The concrete data that instruction is used
}data[2]；
Ing h; This instruction influences the Data Position of other instruction
int?l；
Double ed; Execution result or the intermediateness of this instruction on client computer
}op；
}；
This defined declaration packet comprises data type and concrete data and command status two parts.Data type is represented that by fnc_class table seven has provided the definition to fnc_class; Concrete data are made of jointly ip and op.Ip has provided this machine information, and op has provided concrete instruction or data; Note, in this experimental system, opc[128] definable instruction title; Instruction in this packet comprises the command status composition, i.e. a in the program; Client computer return results or intermediateness information also are the concrete data divisions that has utilized instruction.
Following table seven has provided the definition of fnc_class value:
| The | Meaning | |
| 1 | |
|
| 2 | The operation result of |
|
| 3 | The operation of execution |
|
| 4 | The operation |
|
| 5 | Task termination |
Table seven
3. the realization of instruction and program
Struct id_opc; Order structure
{
char?opc[128]；
struct?DATA
{
char?a；
double?b；
}data[2]；
int?h；
int?l；
double?ed；
}op[NUM]＝
{
" add ", and ' r ＇, 1.0, ＇ r ＇, 2.0,3,1,0.0}; Program
{″dec″，＇r＇，3.0，＇r＇，4.0，3，2，0.0}，
{″mul″，＇w＇，0.0，＇w＇，0.0，4，2，0.0}，
{″div″，＇r＇，7.0，＇w＇，0.0，5，2，0.0}，
{″add″，＇r＇，10.0，＇w＇，0.0，6，1，0.0}，
{″add″，＇w＇，0.0，＇r＇，12.0，7，2，0.0}，
{″add″，＇r＇，2000.0，＇w＇，0.0，9，1，0.0}，
{″add″，＇r＇，10.0，＇r′，20.0，9，2，0.0}，
{″add″，＇w＇，0.0，＇w＇，0.0，10，1，0.0}
}；
In this example, the state in the instruction represents that with a among the structure data its value is that " r " and " w " (meaning is respectively " Ready " and " Wait ", is ready to and waiting status) constitutes.Instruction has four of add (adding), dec (subtracting), mul (taking advantage of), div (removing).The order structure composition that structure in the example is represented is: data mode in instruction title (opc), director data (data), the instruction, (composition in the data structure a), this instructs and dependence (h and the l of other instruction on data to be called for short command status, which bar instruction h represents, which data l represents.Notice that each data all has data mode in the instruction.Ed composition in the structure represents that this instruction deposits the data cell of result and intermediateness after client computer (2) is carried out.
4. a realization flow of main control computer (1) is as shown in Figure 7:
The treatment scheme of main control computer (1) is: initialization system; The instruction client computer packing device (4) of main control computer (1) is obtained the information of this main control computer and is packaged into packet, pass to packet transmitter (12) and broadcast transmission to network, and wait for that other computer system is received the connection request packet of beaming back behind this packet on the network; After the data packet receiver (13) of main control computer (1) is received the connection request packet of other computing machine, send into de-packetizer (5) to its analysis, obtain other computerized information in the packet, this information is sent into client computer round-robin queue in the client computer distributor (14), become a client information; The instruction fetch from program storing unit (3) circularly of the instruction client computer packing device (4) of main control computer (1), from client computer distributor (14), take out the client information of state to be allocated, the two is packaged into packet, pass to packet transmitter (12) and send to client computer (2), wait for the packet that comprises instruction execution result or calculating intermediateness that client computer (2) is beamed back then; The data packet receiver (13) of main control computer (1) enters each subroutine successively and handles respectively after receiving packet from client computer (2).
Described subroutine disposal route comprises:
After data packet receiver (13) is received the packet of client computer (2), handle respectively according to data character in the packet: the packet that receives is sent into de-packetizer (5), de-packetizer (5) is to drawing client information and instruction execution result or calculating intermediateness behind the data packet analysis, client information is sent into client computer round-robin queue in the client computer distributor (14) again, instruction sent in the program storing unit (3) command adapted thereto in the program or data variable group are made amendment, withdraw from subroutine afterwards.
5. corresponding with main control computer (1), a realization flow of client computer (2) as shown in Figure 8:
The disposal route of client computer (2) comprising: client computer (2) system initialization; After the data packet receiver (7) of client computer (2) receives the broadcast message packet of main control computer (1) issue, give reception de-packetizer (9) with data packet delivery, receive de-packetizer (9) and analyze this packet, after obtaining the information of main control computer (1), connect with main control computer (1), after reception packing device (10) is obtained the information of this client computer (2), be packaged into the connection request packet, pass to packet transmitter (8), send to main control computer (1); After main control computer (1) is received, utilize client information and client computer (2) in this connection request packet to connect; After connecting, the packet that main control computer (1) will comprise instruction sends to client computer (2), after client computer (2) is received the packet of main control computer (1), this packet is handled.
Client computer (2) to the data package processing method that comprises instruction that main control computer (1) is sent is: after the data packet receiver (7) of client computer (2) is received packet from main control computer (1), this packet is sent into reception de-packetizer (9), after receiving de-packetizer (9) analysis, instruction execution unit (11) is sent in the instruction that analysis obtains, instruction execution unit (11) is carried out this instruction, instruction execution result or calculating intermediateness are sent into reception packing device (10), receive packing device (10) and be packaged into packet with this client information, pass to packet transmitter (8), send on the network.
6. utilize program in the part 3 in network, to have the data that produce in the implementation under a main control computer and two the client computer situations below, be specifically described.
Main control computer:
1: main control computer sends broadcast message
begain?listenyour?ip?is：211.84.92.92?211.84.92.92?211.84.92.92
2: main control computer and A computer system connect
got?connection?from?211.84.92.90
3, main control computer and B computer system connect
got?connection?from?211.84.92.89
4, the command information of main control computer in A client computer transmission packet
the?data?in?the?package?is：1.000000?2.000000?3?1?0.000000
Illustrate: 1.000000,2.000000 data for the instruction use; The Data Position information of 3 and 1 other instruction that will influence for the result of this instruction, 0.000000 be this instruction on client computer execution result or calculate the intermediateness data bit.
5, the command information of main control computer in B client computer transmission packet
the?data?in?the?package?is：3.000000?4.000000?3?2?0.000000
6, main control computer is received the instruction execution result of A client computer or is calculated intermediateness information
the?result?data?in?the?recvbuf?is：1.000000?2.000000?3?1?3.000000
Illustrate: 1.000000,2.000000 data for the instruction use; The Data Position information of 3 and 1 other instruction that will influence for the result of this instruction; 3.000000 be instruction (add) execution result on the client computer.
7, main control computer inserts the A client information in the client computer round-robin queue again
211.84.92.90
8, main control computer is received the instruction execution result of B computer system or is calculated intermediateness information
the?result?data?in?the?recvbuf?is：3.000000?4.000000?3?2?-1.000000
9, main control computer inserts the B client information in the client computer round-robin queue again
211.84.92.89
Out of Memory repeats no more.
Client computer A:
1, reception is from the broadcast message of main control computer
211.84.92.92
2, send this machine information, connect with main control computer
211.84.92.90
3, reception is from the command information of main control computer packet
pac’s?numble?is：1.000000?2.000000?3?1?0.000000?add
Illustrate: 1.000000,2.000000 data for the instruction use; The Data Position information of 3 and 1 other instruction that will influence for the result of this instruction; 0.000000 be execution result or the calculating intermediateness data bit of this instruction on client computer; Add is the instruction title.
4, send to the command information of the packet of main control computer
pack_ed’s?numble?is：1.000000?2.000000?3?1?3.000000
The relevant information of the out of Memory of client computer A and client computer B repeats no more.
In sum, the present invention utilizes network service, computer system by real work in rational management and the dispatch network, under the prerequisite that does not influence their operate as normal, calculation task with a complexity, be distributed in each computer system of network, make it to obtain parallel processing to a certain extent, like this, by network technology with the numerous computer systems that exist in the network virtual for can parallel processing the based on network dummy multiprocessor system of calculation task, can not only simulate the data processing method of multicomputer system, the speed of handling large-scale user task than existing single processor system improves greatly, and making that system resource has obtained utilizing fully, the real multicomputer system of system cost is low.
Above-mentioned only is specific embodiments of the invention, can not contain technical scope of the present invention, and the present invention should be as the criterion with the described scope of claims.
Claims (9)
1, a kind of based on network dummy multiprocessor system is characterized in that comprising: one and only one with large-scale user task be organized as the program formed by the instruction of based on network dummy multiprocessor system, and with the command assignment of this program to the get on main control computer (1) of processing of other client computer (2); At least one is used for receiving from the instruction of main control computer (1) and handles, returns afterwards result or calculate the client computer (2) that intermediateness is given main control computer (1); Main control computer (1) is connected with computer network with client computer (2).
2, based on network dummy multiprocessor system as claimed in claim 1 is characterized in that: described main control computer (1) comprising: the program storing unit (3) that is used to the program of depositing; Be used for client computer distributor (14) to client computer round-robin queue access client information; Be used for the instruction client computer packing device (4) that instructs and client information is packed and encapsulated; Be used for sending the packet transmitter (12) of packet to client computer (2); Be used to receive data packet receiver (13) from the client data bag; Be used for de-packetizer (5) that the packet that receives is analyzed; Wherein, instruction client computer packing device (4) takes out a client information that is in state to be allocated from client computer distributor (14), from program storing unit (3), get an instruction, the two is packaged into packet, passes to packet transmitter (12) and send to corresponding client computer (2); After data packet receiver (13) receives packet from client computer (2), this packet is sent into de-packetizer (5), and de-packetizer (5) obtains the result of client information and instruction after to this data packet analysis or calculates intermediateness: client information is sent into client computer round-robin queue in the client computer distributor (14) again; The intermediateness of instruction process result or calculating is sent into program storing unit (3), corresponding instruction or data variable group are made amendment.
3, based on network dummy multiprocessor system as claimed in claim 1 is characterized in that: described client computer (2) comprising: be used to receive the data packet receiver (7) from the main control computer packet; Be used to analyze the reception de-packetizer (9) of received data packet; Be used to execute instruction and bear results or calculate the instruction execution unit (11) of intermediateness; Be used for to this machine information with from the result of instruction execution unit (11) or calculate the reception packing device (10) that intermediateness is packaged into the packet that can be received by main control computer (1); Be used for packaged packet is sent to the packet transmitter (8) of main control computer (1); Wherein, after data packet receiver (7) is received the packet that sends from main control computer (1), give reception de-packetizer (9) with this data packet delivery, by receiving de-packetizer (9) it is analyzed, therefrom obtain instruction, instruction execution unit (11) is sent in instruction to be carried out, carry out the back by instruction execution unit (11) result or calculating intermediateness are sent into reception packing device (10), receive packing device (10) with this result or calculate intermediateness and this machine information is packaged into packet, and give packet transmitter (8) with this data packet delivery, send to main control computer (1).
4, a kind of disposal route of based on network dummy multiprocessor is characterized in that: main control computer (1) comprises with the method step that client computer (2) connects:
1), after main control computer (1) start, soon this machine information and send into instruction client computer packing device (4) and be packaged into packet, passes to packet transmitter (12) broadcast transmission to network;
2) after, the data packet receiver of other computer system that exists on the network (7) is received this packet, this packet is sent into reception de-packetizer (9), receiving de-packetizer (9) analyzes this packet and obtain main control computer information in the packet: the main control computer information that obtains is analyzed in this computer system utilization, connect with main control computer (1), and this machine information sent into receive packing device (10) and be packaged into the connection request packet, pass to packet transmitter (8), send to main control computer (1) by packet transmitter (8);
3), the data packet receiver (13) of main control computer (1) is received behind the connection request packet that other computer system is sent, this bag is sent into de-packetizer (5) to be analyzed, the computerized information that de-packetizer (5) obtains analysis deposits the client computer round-robin queue in the client computer distributor (14) in, and the represented computer system of computerized information that enters client computer round-robin queue promptly becomes a client computer (2) of main control computer (1).
5, the disposal route of based on network dummy multiprocessor as claimed in claim 4, described main control computer (1) and client computer (2) are when connecting, and main control computer (1) is characterized in that comprising the following steps: to the storage means of received client information
(1), after main control computer (1) receives the packet of other computer system on the network, this packet is sent into de-packetizer (5), the computerized information that de-packetizer (5) obtains analysis is sent into the client computer round-robin queue of client computer distributor (14);
(2), in client computer round-robin queue, described computerized information is received the precedence composition formation of its packet that sends according to main control computer (1), on article one information of this formation head pointer is set, go out team with the control client information, on the last item information of this formation, tail pointer is set, joins the team with the control client information; The team's action of joining the team out is undertaken by group direction of joining the team out: after the information of head pointer indication was removed, head pointer moved a position backward, points on next bar information position; When a New Client information is sent into, this New Client information is queued at the end of queue, tail pointer is moved a position backward, point on this New Client information position.
6, the disposal route of based on network dummy multiprocessor as claimed in claim 4 is characterized in that: after main control computer (1) and client computer (2) connect, to the process method step of user task be:
Instruction client computer packing device (4) in a, the main control computer (1), from program storing unit (3), take out an instruction, from client computer distributor (14), take out a client information that is in state to be allocated, the two is packaged into packet, sends into packet transmitter (12) and send to client computer (2);
After the data packet receiver (7) of b, client computer (2) is received packet, it is passed to reception de-packetizer (9) parse instruction, sending into instruction execution unit (11) handles, instruction execution unit (11) is sent the result of instruction or the intermediateness of calculating into reception packing device (10), packing with the information of this client computer (2) is packaged into packet, sends into packet transmitter (8) and sends to main control computer (1);
C, after described main control computer (1) receives packet from client computer (2), the packet that receives is sent into de-packetizer (5), after de-packetizer (5) is analyzed, handle respectively according to the information that data packet analysis is obtained: de-packetizer (5) is to obtaining client information and instruction execution result or calculating intermediateness behind the data packet analysis, client information is sent into client computer round-robin queue in the client computer distributor (14) again, instruction execution result or calculating intermediateness are sent in the program storing unit (3), and to wherein command adapted thereto or data variable group are made amendment;
D, enter step a, carry out next step command assignment.
7, the disposal route of based on network dummy multiprocessor as claimed in claim 4, it is characterized in that: the instruction of taking out from described program storing unit (3), this instruction is made up of other director data positional information that instruction title, data, instruction results may influence.
8, the disposal route of based on network dummy multiprocessor as claimed in claim 4 is characterized in that: the instruction of taking out from described program storing unit (3), this instruction also can be made up of instruction title, data variable group.
9, the disposal route of based on network dummy multiprocessor as claimed in claim 7 is characterized in that: described instruction also comprises command status information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005100485179A CN100422935C (en) | 2005-11-08 | 2005-11-08 | Dummy multiprocessor system based on network and processing method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005100485179A CN100422935C (en) | 2005-11-08 | 2005-11-08 | Dummy multiprocessor system based on network and processing method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1963761A true CN1963761A (en) | 2007-05-16 |
| CN100422935C CN100422935C (en) | 2008-10-01 |
Family
ID=38082832
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2005100485179A Expired - Fee Related CN100422935C (en) | 2005-11-08 | 2005-11-08 | Dummy multiprocessor system based on network and processing method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100422935C (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103403688A (en) * | 2011-02-21 | 2013-11-20 | 富士通株式会社 | Processor management method |
| CN113591290A (en) * | 2021-07-20 | 2021-11-02 | 上海华虹宏力半导体制造有限公司 | OPC model simulation method |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| BR112016026264B1 (en) * | 2014-05-09 | 2022-01-18 | Huawei Technologies Co., Ltd | COMPUTER IMPLEMENTED METHOD TO OPTIMIZE A COMPUTER SYSTEM INITIALIZATION TIME AND COMPUTER SYSTEM |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5404550A (en) * | 1991-07-25 | 1995-04-04 | Tandem Computers Incorporated | Method and apparatus for executing tasks by following a linked list of memory packets |
| US6282697B1 (en) * | 1998-09-18 | 2001-08-28 | Wylci Fables | Computer processing and programming method using autonomous data handlers |
| US7249357B2 (en) * | 2001-08-20 | 2007-07-24 | Silicon Graphics, Inc. | Transparent distribution and execution of data in a multiprocessor environment |
-
2005
- 2005-11-08 CN CNB2005100485179A patent/CN100422935C/en not_active Expired - Fee Related
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103403688A (en) * | 2011-02-21 | 2013-11-20 | 富士通株式会社 | Processor management method |
| US9342451B2 (en) | 2011-02-21 | 2016-05-17 | Fujitsu Limited | Processor management method |
| CN113591290A (en) * | 2021-07-20 | 2021-11-02 | 上海华虹宏力半导体制造有限公司 | OPC model simulation method |
| CN113591290B (en) * | 2021-07-20 | 2024-02-06 | 上海华虹宏力半导体制造有限公司 | OPC model simulation method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100422935C (en) | 2008-10-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Enabling in-situ execution of coupled scientific workflow on multi-core platform | |
| Gerbessiotis et al. | Direct bulk-synchronous parallel algorithms | |
| Kruatrachue et al. | Grain size determination for parallel processing | |
| US7647590B2 (en) | Parallel computing system using coordinator and master nodes for load balancing and distributing work | |
| CN105512083B (en) | Method for managing resource, apparatus and system based on YARN | |
| US8676877B2 (en) | Cluster computing using special purpose microprocessors | |
| CN109491795A (en) | Computer cluster for handling calculating task is arranged and its operating method | |
| Amalarethinam et al. | An Overview of the scheduling policies and algorithms in Grid Computing | |
| CN106126323A (en) | Real-time task scheduling method based on cloud platform | |
| CN102968344A (en) | Method for migration scheduling of multiple virtual machines | |
| CN102541858A (en) | Data equality processing method, device and system based on mapping and protocol | |
| CN110050261A (en) | Computational resource allocation | |
| Briat et al. | Athapascan runtime: Efficiency for irregular problems | |
| US20220414437A1 (en) | Parameter caching for neural network accelerators | |
| CN100422935C (en) | Dummy multiprocessor system based on network and processing method | |
| Kim et al. | A parallel and distributed meta-heuristic framework based on partially ordered knowledge sharing | |
| US7454749B2 (en) | Scalable parallel processing on shared memory computers | |
| Temuçin et al. | Efficient multi-path NVLink/PCIe-aware UCX based collective communication for deep learning | |
| Hadadian Nejad Yousefi et al. | IMOS: improved meta-aligner and Minimap2 on spark | |
| EP4232898A1 (en) | Reduction server for fast distributed training | |
| Schreiner et al. | Distributed Maple: parallel computer algebra in networked environments | |
| Böhm et al. | Demaq/Transscale: automated distribution and scalability for declarative applications | |
| Campbell | CLUMPS: a candidate model of efficient, general purpose parallel computation | |
| CN108446165A (en) | A kind of task forecasting method in cloud computing | |
| Faasen | Intermediate uniformly distributed tuple space on transputer meshes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20081001 Termination date: 20101108 |