CN103181132B - Request message processing method and sending method, node and system - Google Patents
Request message processing method and sending method, node and system Download PDFInfo
- Publication number
- CN103181132B CN103181132B CN201280002077.1A CN201280002077A CN103181132B CN 103181132 B CN103181132 B CN 103181132B CN 201280002077 A CN201280002077 A CN 201280002077A CN 103181132 B CN103181132 B CN 103181132B
- Authority
- CN
- China
- Prior art keywords
- request message
- node
- message
- address
- memory address
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 75
- 238000003672 processing method Methods 0.000 title claims abstract description 29
- 230000005540 biological transmission Effects 0.000 claims description 16
- 238000002372 labelling Methods 0.000 claims description 8
- 239000003795 chemical substances by application Substances 0.000 description 10
- 230000006870 function Effects 0.000 description 3
- 108010022579 ATP dependent 26S protease Proteins 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention provides a kind of request message processing method and sending method, node and system.The method comprises: receive the request message that distant-end node sends; Judge whether described request message is winner's message, is, processes described request message, otherwise reject said request message; Wherein, described winner's message refers to that local node can the request message of priority treatment.In embodiment of the present invention technical scheme, node can only process winner's message, then carries out refusal process to other non-winner's messages, thus hardware resource consumption when can reduce node processing request message.
Description
Technical field
The embodiment of the present invention relates to computer technology, particularly relates to request message processing method and sending method, node and system in a kind of concurrent computational system.
Background technology
Along with the development of information technology, the application demand storing for mass data processing, cloud and calculate gets more and more, parallel computing then can well meet such application, wherein, the concurrent computational system adopting cache coherent non-uniform memory access (Cache-CoherentNon-UniformMemoryAccess, CC-NUMA) framework is exactly a kind of conventional parallel computation machine technology.
At present, in non-homogeneous memory access (Non-UniformMemoryAccess, NUMA) framework concurrent computational system, the processor sharing internal memory space of each node, namely the physical memory of each node can by other node visits.Wherein, Node Controller (the NodeController of local node, NC) when receiving and process the multiple request message from distant-end node, usual employing arrives first and first processes, the request message processing method of other requests temporary, this request message that the local agent in the NC of local node (HomeAgent) controller will be made to need buffer memory all, makes Home Agent Director need to have more complicated logic scale; Meanwhile, because a large amount of request message is stored in the buffer memory of Home Agent Director, easily cause local node system bottleneck, even cause deadlock, cause local node to use.
To sum up, the node of existing concurrent computational system adopts to arrive first and first processes, and in the request message processing method of other requests temporary, node needs to store all request messages, and cause the hardware resource consumption of node large, the NC architecture design of node is complicated, and cost is high; Meanwhile, because node resource consumption is large, the dilatation of concurrent computational system is restricted, and node dilatation quantity is restricted, and easily causes system bottleneck.
Summary of the invention
The embodiment of the present invention provides a kind of request message processing method and sending method, node and system, can reduce hardware resource consumption during node processing request message.
First aspect, the embodiment of the present invention provides a kind of request message processing method, comprising:
Receive the request message that distant-end node sends;
Judge whether described request message is winner's message, is, processes described request message, otherwise reject said request message;
Wherein, described winner's message refers to that local node can the request message of priority treatment;
What be rejected maximum request messages in for all request messages of a memory address except winner's message is rejected number of times when being greater than the first predetermined threshold value, described memory address is sent to each distant-end node, notify that described in described each distant-end node, memory address is overheated address, so that described each distant-end node no longer initiates the request message for described memory address.
Second aspect, the embodiment of the present invention also provides a kind of node, comprising:
Request message receiver module, for receiving the request message that distant-end node sends;
Request message processing module, for judging whether described request message is winner's message, is, processes described request message, otherwise reject said request message;
Wherein, described winner's message refers to that local node can the request message of priority treatment;
Overheated address notification module, for be rejected in for all request messages of a memory address except winner's message maximum request messages be rejected number of times be greater than the first predetermined threshold value time, described memory address is sent to each distant-end node, notify that described in described each distant-end node, memory address is overheated address, so that described each distant-end node no longer initiates the request message for described memory address.
The third aspect, the embodiment of the present invention provides a kind of request message treatment system, comprises local node and distant-end node, and described local node is used for processing the request message that described distant-end node sends, and comprises the node that the invention described above embodiment provides.
Fourth aspect, the embodiment of the present invention provides a kind of request message sending method, it is characterized in that, comprising:
When there being request message to send, obtain the memory address that described request message will be asked;
When to inquire described memory address from overheated address table be overheated address, cancel the transmission to described request message, wherein, in described overheated address table, store the memory address of promising overheated address.
5th aspect, the embodiment of the present invention provides a kind of node, comprising:
Memory address acquisition module, for when there being request message to send, obtains the memory address that described request message will be asked;
Request message transmission processing module, when being overheated address for inquiring described memory address from overheated address table, cancels the transmission to described request message, wherein, stores the memory address of promising overheated address in described overheated address table.
The embodiment of the present invention is when processing message, the request message for winner's message to distant-end node sends only is needed to process, refusal process is then carried out to the request message of other non-winner's messages, thus node hardware resources shared when effectively can reduce request message process, reduce the complexity of the NC architecture design of concurrent computational system interior joint, reduce node cost; Simultaneously, node hardware resources shared by process message is few, when the dilatation of concurrent computational system number of nodes, node dilatation quantity will be caused to be restricted because processing request message takies the too much restriction of hardware resource, effectively can improve the convenience of concurrent computational system interior joint dilatation.
Accompanying drawing explanation
In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, be briefly described to the accompanying drawing used required in embodiment or description of the prior art below, apparently, accompanying drawing in the following describes is some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.
The schematic flow sheet of the request message processing method that Fig. 1 provides for the embodiment of the present invention one;
The schematic flow sheet of the request message processing method that Fig. 2 provides for the embodiment of the present invention two;
The schematic flow sheet of the request message processing method that Fig. 3 provides for the embodiment of the present invention three;
The schematic flow sheet of the request message processing method that Fig. 4 provides for the embodiment of the present invention four;
The structural representation of the node that Fig. 5 provides for the embodiment of the present invention five;
The structural representation of the node that Fig. 6 provides for the embodiment of the present invention six;
The structural representation of the node that Fig. 7 provides for the embodiment of the present invention seven;
The structural representation of the node that Fig. 8 provides for the embodiment of the present invention eight;
The structural representation of the node that Fig. 9 provides for the embodiment of the present invention nine;
The structural representation of the request message treatment system that Figure 10 provides for the embodiment of the present invention ten.
Embodiment
For making the object, technical solutions and advantages of the present invention clearly, below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, the every other embodiment that those of ordinary skill in the art obtain under the prerequisite not making creative work, all belongs to the scope of protection of the invention.
The schematic flow sheet of the request message processing method that Fig. 1 provides for the embodiment of the present invention one.The present embodiment can be applicable in concurrent computational system that local node is in the process of the request message of distant-end node, and as shown in Figure 1, the present embodiment request message processing method can comprise the following steps:
Step 101, local node receive the request message that distant-end node sends;
Step 102, local node judge whether this request message is winner's message, is perform step 103, otherwise perform step 104, wherein, this winner's message refers to that local node can the request message of priority treatment;
Step 103, local node process this request message, terminate;
Step 104, refuse this request message.
In the present embodiment, when the local node in concurrent computational system receives the request message of distant-end node transmission, only need to process winner's message, and to non-winner's message, then all carry out refusal process.Like this, local node only needs buffer memory winner message, and the request message not needing buffer memory a large amount of, the hardware resource of node can be saved greatly, the NC framework of node is made only to need less spatial cache can meet process needs to request message, effectively can reduce the complexity of the NC architecture design of node, reduce node hardware cost; Simultaneously, because node only needs to process winner's message, hardware resource takies little, make the number of nodes of concurrent computational system can unrestricted dilatation, and do not worry the problem cannot carrying out System Expansion because node hardware resources is not enough, in addition, node also can be avoided to occur deadlock, the problem that even cannot use because resource consumption is too much.
To sum up, the request message processing method that the embodiment of the present invention provides, the request message for winner's message to distant-end node sends only is needed to process, refusal process is then carried out to the request message of other non-winner's messages, thus node hardware resources shared when effectively can reduce request message process, reduce the complexity of the NC architecture design of concurrent computational system interior joint, reduce node cost; Simultaneously, node hardware resources shared by process message is few, when the dilatation of concurrent computational system number of nodes, node dilatation quantity will be caused to be restricted because processing request message takies the too much restriction of hardware resource, effectively can improve the convenience of concurrent computational system interior joint dilatation.
The schematic flow sheet of the request message processing method that Fig. 2 provides for the embodiment of the present invention two.In the present embodiment, local node can according to request message whether headed by the message that first arrives or determine winner's message according to the refusal number of times of request message, particularly, as shown in Figure 2, the present embodiment request message processing method can comprise the following steps:
Step 201, local node receive the request message that distant-end node sends;
Step 202, judge whether local node has the request message processed, and is, performs step 203, otherwise, perform step 204;
Step 203, local node refuse this message request, and record the unaccepted number of times of this request message, terminate;
Step 204, local node judge the message first arrived headed by this message is whether, are perform step 205, otherwise, perform step 206;
Step 205, this request message is labeled as winner's message, performs step 208;
Step 206, be, when being rejected the maximum request message of number of times, this request message is labeled as winner's message at this request message;
Step 207, judge whether this request message is winner's message, is perform step 208, otherwise, perform step 209;
Step 208, this request message to be processed, terminate;
Step 209, refuse this request message, and record the number of times of this request message, terminate.
In the present embodiment, when local node processes the request message that each distant-end node sends, to can first arrive the request message of local node as winner's message, other request message all can be used as competitor's message, refuse all competitor's messages, and record the unaccepted number of times of each competitor's message; When winner's Message processing is complete, when local node does not have a request message process, can using request messages maximum for refusal number of times as winner's message, when receiving this winner's message next time, can process it.
It will be appreciated by those skilled in the art that; for concurrent computational system; after each distant-end node is rejected to the request message that local node sends; what still can continue sends request message; like this, first send to the request message of local node also usually can have and more be rejected number of times, therefore determine winner's message according to refusal number of times; the requirement message first sent being carried out to priority treatment can be met, thus can avoid the long problem of request message process time delay.
It will be understood by those skilled in the art that in practical application, also can according to the type of request message, and the priority etc. of request message, determine winner's message, or, also can determine winner's message in conjunction with the unaccepted number of times of request message, to this, the embodiment of the present invention is not particularly limited.
It will be appreciated by those skilled in the art that, in practical application, except the maximum message of number of times can will be refused as except winner's message, also can by more than 2 or 2, and refuse the more multiple messages of number of times as winner's message, in practical application, can according to the hardware resource of node, such as, can according to the pipeline depth of local node and cache resources, determine the quantity of winner's message, the present embodiment is to this and be not particularly limited, and processes winner's message of respective numbers as long as local node can meet.
In the present embodiment, the described process to request message, the process of the request message of the same memory address for local node can be referred to, for the process of the request message of other memory addresss for local node, can process according to the processing method with the above-mentioned same memory address for local node.Like this, for the process of the request message for each memory address, all can process according to identical priority, thus the problem that the request message for partial memory address cannot process for a long time can be avoided.
It will be appreciated by those skilled in the art that, in above-mentioned Fig. 2 institute embodiment, what local node stored each request message is rejected number of times, thus can determine whether request message can be winner's message according to the number of times that is rejected of request message, in practical application, local node also can only store be rejected maximum request messages be rejected number of times, after local node receives request message, can according to request message self-contained be rejected number of times, determine whether request message is winner's message, like this, local node only needs less hardware resource to carry out the unaccepted number of times of store and management request message, thus the consumption of hardware resource when can reduce request message process further.Be explained with concrete example below.
The schematic flow sheet of the request message processing method that Fig. 3 provides for the embodiment of the present invention three.Particularly, as shown in Figure 3, the present embodiment request message processing method specifically can comprise the following steps:
Step 301, distant-end node send the request message for local node, and this request message carries the number of times refused by local node;
Step 302, local node receive the request message that distant-end node sends, and judge whether local node has the request message processed, and is, performs step 303, otherwise, perform step 304;
What step 303, local node judged that this request message carries be rejected, and whether number of times equal in all request messages of local node record maximum is rejected number of times, refuse this request message, and maximum refusal number of times of local node record is added 1, terminate, otherwise, refuse this request message, terminate;
Step 304, local node judge the message first arrived headed by this message is whether, are perform step 305, otherwise, perform step 306;
Step 305, this request message is labeled as winner's message, performs step 308;
Step 306, judge that this request message carries be rejected number of times and whether equal the maximum of local node record and be rejected number of times, be that this request message is labeled as winner's message, perform step 308, otherwise, perform step 307;
Step 307, refuse this request message, terminate;
Step 308, the number of times that is rejected maximum in all request messages of local node record to be reset, and this request message is processed, terminate.
In the present embodiment, local node can store be rejected the maximum request message of number of times be rejected number of times, be assumed to be a, and its initial value is set to a=0, like this, after local node receives request message, just a and the self-contained number of times that is rejected of request message can be compared, determine whether request message as winner's message, or can add 1 the need of by a value.
It will be appreciated by those skilled in the art that, that carries in the request message that distant-end node sends is rejected number of times, carry in refusal response message when can be local node refusal, or, respond self statistics by distant-end node according to the refusal that receives, this embodiment of the present invention is not particularly limited.
Can find out, be rejected number of times by carrying in request message in the present embodiment, making local node only need to record maximum refusal number of times, hardware resource consumption during local node processing request message can be reduced further.
At above-mentioned Fig. 1, on Fig. 2 or technical scheme basis embodiment illustrated in fig. 3, local node also can notify that distant-end node stops sending to it request of overheated address, too much to avoid for this overheated Address requests message amount, save broadband resource, particularly, the present embodiment also can comprise the following steps: what be rejected maximum request messages in for other all request messages except winner's message of a memory address is rejected number of times when being greater than the first predetermined threshold value, this memory address is sent to each distant-end node, notify that this memory address of each distant-end node is overheated address, so that each distant-end node no longer initiates the request message for memory address.Further, after memory address for overheated address is sent to distant-end node by local node, what be rejected maximum request messages in for other all request messages except winner's message of this memory address is rejected number of times when being less than the second predetermined threshold value, then this memory address can be sent to each distant-end node, each distant-end node notifies that this memory address of each distant-end node is non-overheat address, so that can initiate the request message for this memory address.Can find out, for time too much to the request refusal number of times of a certain memory address, can notify that distant-end node suspends the request initiated this address, to avoid request message too much and cannot carry out processing and cause concurrent computational system broadband resource to expend excessive problem.
In practical application, distant-end node can safeguard an overheated address table, the memory address for overheated address that each node sends over is stored in this overheated address table, like this, distant-end node, before having new request message to need to send, first can inquire about whether the memory address that its request message that will newly send will ask is overheated address, is stop sending this request message, otherwise carry out poll transmission to it, its specific implementation will be described hereinafter.
It will be understood by those skilled in the art that in practical application, the first above-mentioned predetermined threshold value and the large I of the second predetermined threshold value are set as suitable numerical value as required, such as, can determine according to the network traffic conditions of local node.
In each embodiment of the invention described above, when judging request message as winner's message, first can determine whether the memory address that this request message will be asked can be used, only when available, just it is processed, particularly, the present embodiment also can comprise the following steps: whether the state of inquiring about the memory address that the request message as winner's message will be asked can be used, and when this memory address state is unavailable, refuse this request message, the request of data for this memory address is only just initiated when memory address state is available, request message is processed.In addition, when local node, to inquire the memory address that request message will ask unavailable, such as, just shared by other requests, then can intercept the state of this memory address, until when memory address does not have occupied, just process this request message.
In practical application, the NC of local node can record the catalogue of the state of all memory addresss that it is administered, this catalogue can show whether each memory address takies, namely whether memory address is effective, invalidly be recorded as invalid (invalid) state, otherwise, be recorded as non-disarmed state, and the NC of local node can the state of each memory address of real-time sense, and this catalogue of real-time update.
It will be appreciated by those skilled in the art that, above-mentioned local node and distant-end node all can comprise local agent and caching agent (CacheAgent, CA), wherein, local agent can process the external request message received, and can this request message of buffer memory, the request that CA then will be able to send each core controller (core) in node processes, and is sent to corresponding node.
The schematic flow sheet of the request message processing method that Fig. 4 provides for the embodiment of the present invention four.The present embodiment can be applicable to distant-end node in concurrent computational system to carry out to local node in the transmission of request message, particularly, as shown in Figure 4, when distant-end node needs to send request message, can comprise the steps:
Step 401, when there being request message to send, obtain the memory address that this request message will be asked;
Step 402, when to inquire this memory address be overheated address, cancel the transmission to this request message, wherein, in this overheated address table, store overheated memory address from overheated address table.
In the present embodiment, distant-end node can receive the overheated address that other nodes comprise above-mentioned local node transmission, and be recorded in overheated address table, like this, distant-end node is when sending request message, can arbitrate the request message that the core processor in node will send based on this overheated address table, to determine whether to send request message to destination address, particularly, the request message corresponding with the memory address of the overheated matching addresses in overheated address table can not arbitrated, do not send this request message; To the memory address with the unmatched request in overheated address, (round-robin) fairness doctrine poll can be made to arbitrate, to send request message according to wheel.
In the present embodiment, the request message of arbitration process arrives above-mentioned local node controller, to be processed by local node by internet between sheet.
In the present embodiment, local node, after receiving request message, can process according to scheme shown in Fig. 1 or Fig. 2, to non-winner's message, carries out refusal process, until this request message is set to winner's message.
It will be appreciated by those skilled in the art that, in the present embodiment, local node and distant-end node have included Node Controller and multiple core processor, and local agent and caching agent (CacheAgent), and Node Controller can comprise Home Agent Director and buffer memory agent controller, its concrete structure can have same or similar structure with the node in existing concurrent computational system, do not repeat them here.
The structural representation of the node that Fig. 5 provides for the embodiment of the present invention five.The present embodiment node can be the node in concurrent computational system, and it can realize the method shown in above-mentioned Fig. 1, Fig. 2 or Fig. 3, and particularly, as shown in Figure 5, the present embodiment node can comprise request message receiver module 11 and request message processing module 12, wherein:
Request message receiver module 11, for receiving the request message that distant-end node sends;
Request message processing module 12, for judging whether this request message is winner's message, is, processes this request message, otherwise refuse this request message, wherein, winner's message refers to that local node can the request message of priority treatment.
The present embodiment can process the request message received according to the method step shown in above-mentioned Fig. 1, Fig. 2 or Fig. 3, and its specific implementation see the explanation of the invention described above embodiment of the method one, two or three, can not repeat them here.
It will be understood by those skilled in the art that the present embodiment node is except can realizing the process to request message, also can realize, with other functions of the node in existing concurrent computational system, not repeating them here.
The structural representation of the node that Fig. 6 provides for the embodiment of the present invention six.The embodiment of the present invention can realize above-mentioned method embodiment illustrated in fig. 2, particularly, on above-mentioned technical scheme basis embodiment illustrated in fig. 5, as shown in Figure 6, the present embodiment node also can comprise winner's message judge module 13, for when local node is without the request message processed, judge the message first arrived headed by this request message whether, that this request message is labeled as winner's message, otherwise, if this request message is the request message that unaccepted number of times is maximum, be then winner's message by described request packet labeling.
Further, the request message processing module 12 of the present embodiment node, also for when local node has the request message processed, refusal request message, and the unaccepted number of times of record request message.
In addition, the request message processing module 12 of the present embodiment node, when also can be used for judging request message as winner's message, whether the state of the memory address that inquiry request message will be asked can be used, and when memory address state is unavailable, refuse this request message.
The structural representation of the node that Fig. 7 provides for the embodiment of the present invention seven.In the present embodiment, unaccepted number of times is carried in the request message that distant-end node sends, be rejected number of times at most in all request messages of local node record, particularly, on above-mentioned technical scheme basis embodiment illustrated in fig. 5, as shown in Figure 7, in the present embodiment node, request message processing module 12 can comprise judging unit 121 and the first request message processing unit 122, wherein:
Judging unit 121, for judging whether local node has the request message processed;
First request message processing unit 122, for when local node is without the request message processed, judge whether the unaccepted number of times of this request message that this request message carries equals maximum in all request messages of local node record and be rejected number of times, if, then using request message as winner's message, request message is processed, and the number of times that is rejected maximum in all request messages of local node record is reset, otherwise, refuse this request message.
In addition, as shown in Figure 7, request message processing module 12 also can comprise the second request message processing unit 123, for when local node has the request message processed, what judge that request message carries is rejected number of times and whether equals maximum in all request messages of local node record and be rejected number of times, is maximum number of times that is rejected is added 1, and refuses request message, otherwise, directly refuse request message.
The present embodiment node can realize above-mentioned method embodiment illustrated in fig. 3, and its specific implementation process see the explanation of the invention described above embodiment of the method three, can not repeat them here.
The structural representation of the node that Fig. 8 provides for the embodiment of the present invention eight.On the basis of above-mentioned Fig. 5, Fig. 6 or Fig. 7, as shown in Figure 8, the present embodiment node also can comprise overheated address notification module 15, for be rejected in for the request message of a memory address except winner's message the unaccepted number of times of maximum request messages be greater than the first predetermined threshold value time, this memory address is sent to each distant-end node, notify that memory address described in each distant-end node is overheated address, so that each distant-end node no longer initiates the request message for memory address.And, non-overheat address notification module 16, when unaccepted number of times for being rejected maximum request messages in for other request messages except winner's message of this memory address is less than the second predetermined threshold value, this memory address is sent to each distant-end node, each distant-end node notifies that this memory address of each distant-end node is non-overheat address, so that can initiate the request message for memory address.
The structural representation of the node that Fig. 9 provides for the embodiment of the present invention nine.The present embodiment node can be each node of concurrent computational system, can send request message to other nodes, and particularly, as shown in Figure 9, the present embodiment node can comprise memory address acquisition module 21 and request message transmission processing module 22, wherein:
Memory address acquisition module 21, for when there being request message to send, obtains the memory address that this request message will be asked;
Request message transmission processing module 22, when being overheated address for inquiring this memory address from overheated address table, cancels the transmission to this request message, wherein, stores overheated memory address in this overheated address table.
As shown in Figure 9, the present embodiment node also can comprise overheated address accept processing module 23, for receiving the memory address for overheated address that other nodes send, and is added in overheated address table by the memory address for overheated address received.And, non-overheat address accept processing module 24, for receiving the memory address for non-overheat address that other nodes send, and the memory address that will delete from overheated address table as non-overheat address.
The present embodiment node is before carrying out request message transmission, the request message for overheated address can be filtered out, only carry out arbitration to the request message of non-overheat address to send, like this, the transmission of a large amount of not treatable request message can be avoided, can reduce the node of request message reception to the process of request message, its specific implementation see the explanation of embodiment of the method shown in above-mentioned Fig. 4, can not repeat them here.
The structural representation of the request message treatment system that Figure 10 provides for the embodiment of the present invention ten.As shown in Figure 10, the present embodiment request message treatment system is concurrent computational system, comprise local node 10 and distant-end node 20, the request message that this local node 10 can be used for distant-end node 20 sends processes, specifically can comprise the node shown in Fig. 5, Fig. 6, Fig. 7 or Fig. 8, its concrete structure and function see the explanation of above-mentioned Fig. 5, Fig. 6, Fig. 7 or Fig. 8, can not repeat them here.
In the present embodiment, above-mentioned distant-end node 20 can comprise the node shown in above-mentioned Fig. 9, and its concrete structure see shown in above-mentioned Fig. 9, can not repeat them here.
In the present embodiment, local node 10 and distant-end node 20 all can comprise the node of above-mentioned Fig. 5, Fig. 6, Fig. 7 or Fig. 8, and the node shown in Fig. 9, namely local node 10 and distant-end node 20 can have identical 26S Proteasome Structure and Function, like this, local node and distant-end node all can receive the request message that other nodes send, and also can send request message to other nodes simultaneously, its specific implementation see the explanation of the invention described above embodiment of the method and device embodiment, can not repeat them here.
Those skilled in the art can be well understood to, and for convenience and simplicity of description, the system of foregoing description and the specific works process of node, with reference to the corresponding process in preceding method embodiment, can not repeat them here.
One of ordinary skill in the art will appreciate that: all or part of step realizing said method embodiment can have been come by the hardware that program command is relevant, aforesaid program can be stored in a computer read/write memory medium, this program, when performing, performs the step comprising said method embodiment; And aforesaid storage medium comprises: ROM, RAM, magnetic disc or CD etc. various can be program code stored medium.
Last it is noted that above each embodiment is only in order to illustrate technical scheme of the present invention, be not intended to limit; Although with reference to foregoing embodiments to invention has been detailed description, those of ordinary skill in the art is to be understood that: it still can be modified to the technical scheme described in foregoing embodiments, or carries out equivalent replacement to wherein some or all of technical characteristic; And these amendments or replacement, do not make the essence of appropriate technical solution depart from the scope of various embodiments of the present invention technical scheme.
Claims (22)
1. a request message processing method, is characterized in that, comprising:
Receive the request message that distant-end node sends;
Judge whether described request message is winner's message, is, processes described request message, otherwise reject said request message; Wherein, described winner's message refers to that local node can the request message of priority treatment;
What be rejected maximum request messages in for all request messages of a memory address except winner's message is rejected number of times when being greater than the first predetermined threshold value, described memory address is sent to each distant-end node, notify that described in described each distant-end node, memory address is overheated address, so that described each distant-end node no longer initiates the request message for described memory address.
2. request message processing method according to claim 1, is characterized in that, described judge whether described request message is winner's message before, also comprise:
When described local node is without the request message processed, judge the message first arrived headed by described request message whether, be then winner's message by described request packet labeling, otherwise, if described request message is be rejected the maximum request message of number of times, be then winner's message by described request packet labeling.
3. request message processing method according to claim 2, is characterized in that, after the request message that described reception distant-end node sends, also comprises:
When described local node has the request message processed, reject said request message, and record the unaccepted number of times of described request message.
4. request message processing method according to claim 3, is characterized in that, if described described request message is described request message be rejected the maximum request message of number of times, is then winner's message by described request packet labeling, comprises:
Judge whether the unaccepted number of times of described request message that described request message carries equals maximum in all request messages of described local node record and be rejected number of times, if, be then winner's message by described request packet labeling, and the number of times that is rejected maximum in all request messages of described local node record is reset.
5. request message processing method according to claim 4, is characterized in that, also comprises:
When described local node has the request message processed, what judge that described request message carries is rejected number of times and whether equals maximum in all request messages of described local node record and be rejected number of times, is that the described number of times that is rejected at most is added 1.
6., according to the arbitrary described request message processing method of claim 1-5, it is characterized in that, described described memory address is sent to each distant-end node after, also comprise:
What be rejected maximum request messages in for all request messages of described memory address except winner's message is rejected number of times when being less than the second predetermined threshold value, described memory address is sent to each distant-end node, notify that described in described each distant-end node, memory address is non-overheat address, so that described each distant-end node can initiate the request message for described memory address, described second predetermined threshold value is less than described first predetermined threshold value.
7. request message processing method according to claim 1, is characterized in that, before processing described request message, also comprises:
Whether the state of the memory address that inquiry described request message will be asked can be used, and when described memory address state is unavailable, reject said request message.
8. request message processing method according to claim 1, is characterized in that, described method is carry out for the request message of the same memory address of local node the method that processes.
9. a node, is characterized in that, comprising:
Request message receiver module, for receiving the request message that distant-end node sends;
Request message processing module, for judging whether described request message is winner's message, is, processes described request message, otherwise reject said request message; Wherein, described winner's message refers to that local node can the request message of priority treatment;
Overheated address notification module, for be rejected in for all request messages of a memory address except winner's message maximum request messages be rejected number of times be greater than the first predetermined threshold value time, described memory address is sent to each distant-end node, notify that described in described each distant-end node, memory address is overheated address, so that described each distant-end node no longer initiates the request message for described memory address.
10. node according to claim 9, is characterized in that, also comprises:
Winner's message judge module, for when described local node is without the request message processed, judge the message first arrived headed by described request message whether, be then winner's message by described request packet labeling, otherwise, if described request message is be rejected the maximum request message of number of times, be then winner's message by described request packet labeling.
11. nodes according to claim 10, is characterized in that, described request message processing module (MPM), and also for when described local node has the request message processed, reject said request message, and record the unaccepted number of times of described request message.
12. nodes according to claim 11, is characterized in that, described request message processing module (MPM) comprises:
Judging unit, for judging whether described local node has the request message processed;
First request message processing unit, for when described local node is without the request message processed, judge whether the unaccepted number of times of described request message that described request message carries equals maximum in all request messages of described local node record and be rejected number of times, if, be then winner's message by described request packet labeling, described request message is processed, and the number of times that is rejected maximum in all request messages of described local node record is reset, otherwise reject said request message.
13. nodes according to claim 12, is characterized in that, described request message processing module (MPM) also comprises:
Second request message processing unit, for when described local node has the request message processed, what judge that described request message carries is rejected number of times and whether equals maximum in all request messages of described local node record and be rejected number of times, that the described number of times that is rejected at most is added 1, and the message that rejects said request, otherwise reject said request message.
14., according to the arbitrary described node of claim 9-13, is characterized in that, also comprise:
Non-overheat address notification module, for be rejected in for all request messages of described memory address except winner's message maximum request messages be rejected number of times be less than the second predetermined threshold value time, described memory address is sent to each distant-end node, notify that described in described each distant-end node, memory address is non-overheat address, so that described each distant-end node can initiate the request message for described memory address, described second predetermined threshold value is less than described first predetermined threshold value.
15. nodes according to claim 9, it is characterized in that, described request message processing module (MPM), time also for judging described request message as winner's message, whether the state of the memory address that inquiry described request message will be asked can be used, and when described memory address state is unavailable, reject said request message.
16. 1 kinds of request message sending methods, is characterized in that, comprising:
When there being request message to send, obtain the memory address that described request message will be asked;
When to inquire described memory address from overheated address table be overheated address, cancel the transmission to described request message, wherein, in described overheated address table, store the memory address for overheated address that each node sends over; The memory address for overheated address that node described in each sends over meets: what receive at described node comprises in the request message of this memory address, the number of times that is rejected being rejected maximum request messages in other all request messages except winner's message is greater than first threshold, described winner's message for node described in each can the request message of priority treatment.
17. request message sending methods according to claim 16, is characterized in that, also comprise:
Receive the memory address for overheated address that other nodes send, and the memory address described in reception being overheated address is added in described overheated address table.
18. request message sending methods according to claim 16 or 17, is characterized in that, also comprise:
Receive the memory address for non-overheat address that other nodes send, and be the memory address of non-overheat address described in deleting from described overheated address table.
19. 1 kinds of nodes, is characterized in that, comprising:
Memory address acquisition module, for when there being request message to send, obtains the memory address that described request message will be asked;
Request message transmission processing module, when being overheated address for inquiring described memory address from overheated address table, cancel the transmission to described request message, wherein, in described overheated address table, store the memory address for overheated address that each node sends over; The memory address for overheated address that node described in each sends over meets: what receive at described node comprises in the request message of this memory address, the number of times that is rejected being rejected maximum request messages in other all request messages except winner's message is greater than first threshold, described winner's message for node described in each can the request message of priority treatment.
20. nodes according to claim 19, is characterized in that, also comprise:
Overheated address accept processing module, for receiving the memory address for overheated address that other nodes send, and is added in the memory address described in reception being overheated address in described overheated address table.
21. nodes according to claim 19 or 20, is characterized in that, also comprise:
Non-overheat address accept processing module for receiving the memory address for non-overheat address that other nodes send, and is the memory address of non-overheat address described in deleting from described overheated address table.
22. 1 kinds of request message treatment systems, is characterized in that, comprise local node and distant-end node, and described local node is used for processing the request message that described distant-end node sends, and comprise the arbitrary described node of claim 9-15; Described distant-end node comprises the arbitrary described node of claim 19-21.
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2012/082955 WO2014059576A1 (en) | 2012-10-15 | 2012-10-15 | Method for processing request message, sending method, node and system thereof |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103181132A CN103181132A (en) | 2013-06-26 |
CN103181132B true CN103181132B (en) | 2015-11-25 |
Family
ID=48639400
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201280002077.1A Active CN103181132B (en) | 2012-10-15 | 2012-10-15 | Request message processing method and sending method, node and system |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN103181132B (en) |
WO (1) | WO2014059576A1 (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105812266B (en) * | 2014-12-31 | 2018-10-23 | 北京东土科技股份有限公司 | A kind of hardware configuration processing method and processing device of request message |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101635678A (en) * | 2009-06-15 | 2010-01-27 | 中兴通讯股份有限公司 | Method and system for controlling P2P terminal flow |
CN101908953A (en) * | 2009-06-02 | 2010-12-08 | 中兴通讯股份有限公司 | Method and device for dispatching retransmission data |
CN102594691A (en) * | 2012-02-23 | 2012-07-18 | 中兴通讯股份有限公司 | Method and device for processing message |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6266743B1 (en) * | 1999-02-26 | 2001-07-24 | International Business Machines Corporation | Method and system for providing an eviction protocol within a non-uniform memory access system |
US7305492B2 (en) * | 2001-07-06 | 2007-12-04 | Juniper Networks, Inc. | Content service aggregation system |
WO2012119369A1 (en) * | 2011-08-02 | 2012-09-13 | 华为技术有限公司 | Message processing method, device and system based on cc-numa |
CN102594669A (en) * | 2012-02-06 | 2012-07-18 | 福建星网锐捷网络有限公司 | Data message processing method, device and equipment |
-
2012
- 2012-10-15 CN CN201280002077.1A patent/CN103181132B/en active Active
- 2012-10-15 WO PCT/CN2012/082955 patent/WO2014059576A1/en active Application Filing
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101908953A (en) * | 2009-06-02 | 2010-12-08 | 中兴通讯股份有限公司 | Method and device for dispatching retransmission data |
CN101635678A (en) * | 2009-06-15 | 2010-01-27 | 中兴通讯股份有限公司 | Method and system for controlling P2P terminal flow |
CN102594691A (en) * | 2012-02-23 | 2012-07-18 | 中兴通讯股份有限公司 | Method and device for processing message |
Also Published As
Publication number | Publication date |
---|---|
CN103181132A (en) | 2013-06-26 |
WO2014059576A1 (en) | 2014-04-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8209690B2 (en) | System and method for thread handling in multithreaded parallel computing of nested threads | |
US8250164B2 (en) | Query performance data on parallel computer system having compute nodes | |
TWI519958B (en) | Method and apparatus for memory allocation in a multi-node system | |
US7155722B1 (en) | System and method for process load balancing in a multi-processor environment | |
CN101616083B (en) | Message forwarding method and device | |
CN106936739B (en) | Message forwarding method and device | |
CN102843426B (en) | Based on Web cache resources shared system and the method for intelligent father node | |
US20050038941A1 (en) | Method and apparatus for accessing a memory | |
US20090307408A1 (en) | Peer-to-Peer Embedded System Communication Method and Apparatus | |
TW201543218A (en) | Chip device and method for multi-core network processor interconnect with multi-node connection | |
CN110119304B (en) | Interrupt processing method and device and server | |
CN104657308A (en) | Method for realizing server hardware acceleration by using FPGA | |
CN109697122A (en) | Task processing method, equipment and computer storage medium | |
CN107562685B (en) | Method for data interaction between multi-core processor cores based on delay compensation | |
US11231964B2 (en) | Computing device shared resource lock allocation | |
US20240205170A1 (en) | Communication method based on user-mode protocol stack, and corresponding apparatus | |
CN111404931A (en) | Remote data transmission method based on persistent memory | |
CN105302489A (en) | Heterogeneous multi-core remote embedded memory system and method | |
CN111240853A (en) | Method and system for bidirectionally transmitting massive data in node | |
CN117076140B (en) | Distributed computing method, device, equipment, system and readable storage medium | |
CN115114042A (en) | Storage data access method and device, electronic equipment and storage medium | |
CN103181132B (en) | Request message processing method and sending method, node and system | |
CN112596669A (en) | Data processing method and device based on distributed storage | |
CN104125292A (en) | Data processing device, cloud server and use method thereof | |
CN111190840A (en) | Multi-party central processing unit communication architecture based on field programmable gate array control |
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 | ||
TR01 | Transfer of patent right |
Effective date of registration: 20211224 Address after: 450046 Floor 9, building 1, Zhengshang Boya Plaza, Longzihu wisdom Island, Zhengdong New Area, Zhengzhou City, Henan Province Patentee after: xFusion Digital Technologies Co., Ltd. Address before: 518129 Bantian HUAWEI headquarters office building, Longgang District, Guangdong, Shenzhen Patentee before: HUAWEI TECHNOLOGIES Co.,Ltd. |
|
TR01 | Transfer of patent right |