CN101866303B - High-performance message transmission system with low latency time - Google Patents
High-performance message transmission system with low latency time Download PDFInfo
- Publication number
- CN101866303B CN101866303B CN201010164028.0A CN201010164028A CN101866303B CN 101866303 B CN101866303 B CN 101866303B CN 201010164028 A CN201010164028 A CN 201010164028A CN 101866303 B CN101866303 B CN 101866303B
- Authority
- CN
- China
- Prior art keywords
- store buffer
- queue
- message
- buffer district
- pointed
- 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
Abstract
The present invention relates to a high-performance message transmission system with a low latency time. Disclosed are a system and a method of high-performance message transmission with a low latency time among at least two courses on at least one node. The system includes a storage device (100) for storing course-between message in a queue (102) of a storage buffer area (104), and the storage buffer area (104) is arranged in a share memory accessed by multiple courses in parallel. Pointing units (106, 108) point to the queue (102) to insert the message to the queue or fetch the message from the queue (102). The course inserts and fetches the message by a unlocking technique, such that system throughput is increased and the system is adapted to be implemented by applications, e.g. finance business.
Description
Technical field
The present invention relates to the communications field.
Especially, the present invention relates to the field of carrying out communicating between two or more distributed computers.
Background technology
In the calculation, application is provided, one of the most critical factor of summary responses of the high speed operation of software and Web application is message transfer service.
The performance using efficient message transfer service can significantly improve software and the application using this system also transfers the overall performance improving computing machine.
Interprocess communication between the process that message transfer service relates to for the operation room of each process of association and each application substantially.Message transfer service transmission package contains the message of the data that will exchange between process, speed-up computation process thus.
The various message transfer services of current use have the queue of locking, and the queue of described locking, for storing the message from each process, consequently has and operates shortcoming slowly.
No. 6029205th, United States Patent (USP) discloses the various system for shared storage interprocess communication and the technology of pass-along message between the process performed, but it does not overcome the problem of locking queue.
U.S. Patent application 20050021622 discloses a kind of at publisher node with order the system of carrying out dynamic message route between node for theme, it comprises multiple message queue, at least one theme node table, orders module, release module and other between one or more publisher and one or more ordering party, send the module of message.This document is only absorbed in and is issued the mechanism of order and inapplicable when point-to-point queue.This application does not overcome the problem of locking queue yet.
No. 6029205th, United States Patent (USP) discloses a kind of system and method just carrying out interprocess communication between the sequential process of the cooperation of executed in parallel, and it uses shared storage queue as the mechanism for Message Transmission and Process Synchronization.Message stores is the new list item (enqueue operations) in queue by transmission process, and described list item is visible in the virtual address space of the first process, and receiving process reads message from queue and performs out team's operation.But it does not specifically mention/queue of claimed use shared storage reaches low latency.
Therefore, need a kind of system and method, it:
● multiple process can be communicated simultaneously;
● improve the handling capacity of message transfer service;
● be easy to be deployed on existing system;
● need the infrastructure of minimum level; And
● for key application provides efficient message transferring technology.
Summary of the invention
A target of the present invention is to provide a kind of message transfer service efficiently and reduces computing time thus.
Another target of the present invention is to provide a kind of message transfer service of cost-effective.
Another target of the present invention is to provide a kind of system that multiple process can be communicated simultaneously.
Another target of the present invention is to provide a kind of system providing high-throughput and low latency.
Another target of the present invention is to provide the system together with a kind of can being easily deployed in existing system.
Another target of the present invention is to provide a kind of system needing the infrastructure of minimum level.
Another target of the present invention is to provide a kind of efficient message transferring technology for key application.
And then a target of the present invention also can use in the hardware and software infrastructure with poor memory consistency model.
Invention contemplates the system of carrying out message transmission in the interprocess communication between a kind of at least two processes for running at least one node, described system comprises:
● storage facilities, is suitable for message between storage process;
● store buffer sector arranges, in the shared storage can accessed by multiple task parallelism of described store buffer district in described storage facilities;
● idle sensing unit, with the process context being suitable for the idle store buffer district pointed in described queue;
● at least one data points to unit, with the process context being suitable for pointing to the store buffer district comprising inter-process messages; And
● message read/write device, be suitable for: when not locking described queue, by allowing process, message is inserted by the described idle described idle store buffer district pointed to pointed by unit, and allow at least one process to read to be stored in comprise the insertion message pointed to by described data in the store buffer district of the inter-process messages pointed by unit, insert from the described store buffer district described queue simultaneously and take out message.
Typically, described system comprises updating device, is suitable for upgrading the position that the described free time points to unit and described data sensing unit.
Typically, described shared storage is positioned at local node or remote node.In addition, described process can be run at remote node place, described remote node runs and uses communicator to carry out the process communicated, and described communicator is suitable for promoting the connection from comprising the group selection that TCP/IP connects, GPRS connects, WiFi connects, WiMax connects and EDGE connects.
Described read/write device comprises testing fixture, for write/update, whether described testing fixture checks that described queue is whether full and be empty for reading/take out queue described in operation inspection.
According to a preferred embodiment of the invention, described queue comprises hereof.
Preferably, described file is mapped in the main memory space of processor.
Again, according to a preferred embodiment of the invention, each store buffer district in queue is linked to next store buffer district and last store buffer district is linked to first store buffer district, forms the sequence of circular linkage thus to be stored in one by one in described store buffer district by all inbound messages.
This system is preferably financial transaction system, but also may be used for other application many.
According to a further aspect in the invention, providing the method for carrying out message transmission in the interprocess communication between a kind of at least two processes for running at least one node, said method comprising the steps of:
● provide storage facilities with message between storage process;
● be provided at least one the store buffer sector row in the shared storage in described storage facilities;
● provide idle and point to unit, its process context with the idle store buffer district pointed in described queue;
● provide at least one data to point to unit, it comprises the process context in the store buffer district of described inter-process messages with pointing to;
● receive at least one inter-process messages from least one process run at least one node;
● received inter-process messages are inserted the store buffer district pointed by described free time sensing unit of described queue;
● the location updating being pointed to unit the free time is that next idle store buffer district of described queue is to hold next inter-process messages;
● take out inter-process messages by multiple process from the store buffer district pointed to pointed by unit by the corresponding data of described multiple process of described queue simultaneously; And
● the location updating described data being pointed to unit is comprise next the store buffer district by the message read by arbitrary described process.
Typically, the idle step pointing to the position of unit of renewal comprises the following steps: check whether described free time sensing unit points to the buffer zone pointed by arbitrary data sensing unit, if, then wait for that this buffer zone becomes the free time, otherwise the pointer position being pointed to unit the free time is updated to next idle store buffer district.And the step of taking out inter-process messages comprises the following steps: check that described data are pointed to unit and whether pointed to by the described idle buffer zone pointed to pointed by unit, if so, then wait for and data are inserted this queue to take out.
Typically, the step inserting the inter-process messages received comprises the following steps: by inserted message copy to the primary memory of described node so that permanent/secondary stores.
Accompanying drawing explanation
Referring now to accompanying drawing, the present invention is described, these accompanying drawings are:
Fig. 1 shows the system for carrying out in interprocess communication without lock message transmission (locklessmessaging);
Fig. 2 shows the detailed view of the queue of the storage unit being a series of circular linkage; And
Fig. 3 shows for carrying out the method transmitted without lock message in interprocess communication.
Embodiment
Prior art does not provide the system and method without lock Message Transmission any platform having poor memory consistency model carrying out inter-process messages.Therefore, even if the invention provides the efficient system and method that a kind of platform having poor memory consistency model also can carry out in interprocess communication nothing lock Message Transmission.
Efficient interprocess communication is realized between the process of the message transfer service that the present invention conceives between the process of same server (local message transmission) and on different server (remote server).
According to embodiments of the invention, the message transfer service for interprocess communication realizes hereof.These files comprise message queue, that is, the data structure of message between storage process.These files are mapped to the main memory space of processor to accelerate access.
Described queue comprises store buffer district, and each store buffer district is linked to next store buffer district and last store buffer district is linked to first store buffer district, form the sequence of circular linkage thus to be stored in one by one in described store buffer district by all inbound messages.In addition, described queue has the sensing unit of two types, and namely idle sensing unit and data point to unit.
Described queue has idle points to unit, the described free time point to unit point to the idle store buffer district that wherein can store next inbound messages all the time and all the time with the process context writing described queue.
Described queue has at least one data and points to unit, described data point to unit point to all the time can from next store buffer district of the message of its reading pre-stored and all the time with intend to carry out from described queue the process context that reads.When multiple process reads from same queue, multiple data can be adopted to point to unit for reading message from described queue, wherein each data are pointed to unit and are read process context with one.
In addition, idle end points to unit and one or more data point to the position of unit for checking whether completely whether queue be empty and queue sky and all non-storing message in arbitrary store buffer district, all stored the message will read by one or more process in all store buffer districts of completely i.e. queue.
Data are only had to point to unit (namely in queue, queue is only read with one and point-to-point queue that one writes process context) when, full by inspection queue or for empty, mode is: determine that idle unit and the data pointed to are pointed to the position of unit and follow the process of the following stated:
● if data point to unit and the idle unit that points to points to same store buffer district, then queue is empty.
● if the idle store buffer district pointed to pointed by unit makes its next buffer zone be pointed to by data anchor (Data Anchor), then queue is full.In the case, the node pointed by free pointer is next enabled node.
When queue has multiple data sensing unit, namely described queue orders queue (publish subscribequeue) with some issues of reading reading side's process context of message from same queue, and each process has independent data points to unit, and at any one time, only have a process can by self and the idle unit that points to being associated and write queue.Check that this Class Queue is full or for empty, mode is: determine that idle unit and all data pointed to are pointed to the position of unit and follow the process of the following stated:
● if all data point to unit and the idle unit that points to all points to same store buffer district, then queue is empty.
● if the idle store buffer district pointed to pointed by unit makes its next buffer zone be pointed to by arbitrary data sensing unit, then queue is full.In the case, the node pointed by free pointer is next enabled node.
According to the present invention, the file of memory mapped can be stored in local machine, namely in the node running multiple process or stand-alone computer, local machine also adopts message queue and hereof by File Mapping in the primary memory of processor, to promote the interprocess communication between the process run on same stand-alone computer, mode is: between any two processes, use point-to-point queue, or queue is ordered in the issue using wherein one or more processes to read message from single queue.
In addition, according to the present invention, the file of independent memory mapped can be stored in remote machine, namely, can be coupled to each other with any node in the node of the long-range existence forming network communicatedly, to promote the interprocess communication between the process run on remote node, mode is: use point-to-point queue between any two processes run on remote node, or queue is ordered in the issue using the one or more processes wherein run on remote node to read message from single queue.
In these cases, write side's process that a certain node of network runs by message stores in its message queue that have mapped in the shared storage of message queue file, and then, this message is read by middle transmit leg process, then this message is sent to the Section Point running another middle take over party's process by middle transmit leg process by network, another middle take over party's process described reads message and is stored in the message queue of the shared storage being mapped to this node, reads process read for reality.
According to the present invention, message can be write queue when the locking without any type simultaneously and read message from queue.
As described in the prior art, the side's of write process will lock queue during the write operation to queue, and make until write operation terminates, other processes just can read; And the side's of reading process will lock queue to the read operations of queue, and make until read operation terminates, other processes could write queue, this will cause low-down handling capacity and higher stand-by period.Together with use lock, prior art process is also at the condition place time-out that queue is empty or full.This time-out process and the operation making process return execution bring a large amount of expense and increase the average latency.
In order to eliminate this expense, it is empty or full for the present invention proposes continuous review queue.This contributes to keeping the low-down stand-by period.
In order to there is this type of without lock message transfer service, below list the some constraints will applied equally:
● the side's of write process can only write queue point to store buffer district pointed by unit by the free time and the side's of reading process can only read and points to store buffer district pointed by unit by data.
● processor or chipset or operating system should not revise the order of the instruction of process for optimizing object, this can cause the handling capacity reduced.
● after process has read or write queue, only just can carry out the idle renewal pointing to unit and data sensing unit.
● use instruction and compiler directive (such as, volatile key word (volatile key), memory barrier instruction) to guarantee that message queueing correctly works in the Hardware-software platform with poor memory consistency.
According to a preferred embodiment of the invention, advantageously message transfer service can be adopted to promote the assets financial transaction across continent of negotiator, middle man, businesses and institutions with the local and remote interprocess communication stand-by period of very high speed and reduction.
With reference to figure 1, show for carrying out the system transmitted without lock message in the interprocess communication between at least two processes.Described process an independently Nodes operation, or can different nodes coupled to each other communicatedly can be run in a network.
Numeral 100 in node A represents storage facilities, and it safeguards the queue 102 in store buffer district 104, and the latter stores will from a process transmission to the inter-process messages of another process.
As shown in Figure 2, queue is according to a preferred embodiment of the invention the store buffer district of a series of circular linkage, that is, each store buffer district points to next the store buffer district in series and last store buffer district points to first store buffer district to form circulation.
Queue 102 can be different point-to-point queues or issue order queue, and therefrom multiple reading sides process can use the specific data of their process to point to unit to read message.
When communicating between the local process run all on the same node with the side's of write process in reading side, the side's of write process is by the store buffer district 104 pointed by free time sensing unit 106 of message input queue 12.After write store buffer district 104, upgrade the idle position pointing to unit 106, and it points to next idle store buffer district 104 that wherein can write another message of queue.
And then, if queue is point-to-point type, arbitrary process in two processes associated with this queue all can write message in the current store buffer district 104 pointed to by the free time pointed by unit 106, and order queue for issue, any process by queue message transfer all can write current store buffer district 104 pointed by free time sensing unit 106.
The side's of reading process reads in the message stored in the store buffer district 104 pointed by data sensing unit of queue 102.Once message is stored in store buffer district 104, the side's of reading process just reads this store buffer district.Correspondingly, more new data points to the position of unit to point to next the store buffer district 104 comprising next message that will read.
When point-to-point queue, arbitrary process in two association process all can read the message stored in the store buffer district 104 pointed by data sensing unit 108 of this queue, and when issuing subscribe message and transmitting, the arbitrary process transmitted by this queue in the process of message all can read message.Difference is herein that each process will have independent data and point to unit, and the process that message is read in any hope all points to message pointed by unit 108 by reading by its oneself data.
Fig. 1 also show two nodes connected in a network, i.e. node A and Node B, and they can run difference mutual each other and read and write process.In an identical manner, the connection that TCP/IP connects or any other is applicable to is used to make multiple node interconnected amongst one another, so that the system and method using the present invention to conceive runs multiple process to carry out interprocess communication also within the scope of the invention.
With reference to figure 3, showing a kind of for carrying out the method transmitted without lock message in the interprocess communication between at least two processes, said method comprising the steps of:
● provide storage facilities with message between storage process, 1001;
● be provided in the store buffer sector row in the shared storage in described storage facilities, 1003;
● provide idle and point to unit, its process context with the idle store buffer district pointed in described queue, 1005;
● provide at least one data to point to unit, it comprises the process context in the store buffer district of described inter-process messages with pointing to, and 1007;
● receive at least one inter-process messages, 1009 from least one process run at least one node;
● received inter-process messages are inserted the store buffer district pointed by described free time sensing unit of described queue, 1011;
● the location updating being pointed to unit the free time is that next idle store buffer district of described queue is to hold next inter-process messages, 1013;
● take out inter-process messages, 1015 by multiple process from the store buffer district pointed to pointed by unit by the corresponding data of described multiple process of described queue simultaneously; And
● the location updating described data being pointed to unit is comprise next the store buffer district by the message read by arbitrary described process, 1017.
In addition, instruction and compiler directive is used with predefined procedure, such as volatile key word and memory barrier instruction, to guarantee, after the step in the store buffer district pointed by described free time sensing unit received inter-process messages being inserted described queue, the content in store buffer district to be brought down stocks (flush) to primary memory (RAM) from high-speed cache.
Technical progress
Technical progress of the present invention comprises and providing:
● a kind of efficient message transfer service reducing computing time.
● a kind of message transfer service of cost-effective.
● a kind of system that multiple process can be communicated simultaneously.
● a kind of system that high-throughput and low latency are provided.
● a kind of system being easy to dispose together with existing system.
● a kind of system needing the infrastructure of minimum level.
● a kind of efficient message transferring technology for key application.
Although been considerably highlight assembly and the components of preferred embodiment at this, will be appreciated that and can conceive many embodiments and can many changes be made in a preferred embodiment and not depart from principle of the present invention.According to disclosure herein, these and other changes in preferred embodiment and other embodiments of the present invention will be apparent to those skilled in the art, to clearly understand thus, foregoing description content should be interpreted as just illustrating the present invention and unrestricted the present invention.
Claims (16)
1. carry out a system for message transmission in the interprocess communication between at least two processes run at least one node, described system comprises:
Storage facilities, is suitable for message between storage process;
Store buffer sector arranges, in the shared storage can accessed by multiple task parallelism of described store buffer district in described storage facilities;
Idle sensing unit, with the write process context being suitable for the idle store buffer district pointed in described queue;
At least one data points to unit, with the reading process context being suitable for pointing to the store buffer district comprising inter-process messages; And
Message read/write device, be suitable for: when not locking described queue, by allowing said write process message being inserted the insertion message pointed to the described idle store buffer district pointed by unit by the described free time and allow described reading process to read and be stored in and comprise and to be pointed to by described data in the store buffer district of the inter-process messages pointed by unit, inserting from the described store buffer district described queue simultaneously and taking out message.
2. the system as claimed in claim 1, described system comprises updating device, is suitable for upgrading the position that the described free time points to unit and described data sensing unit.
3. the system as claimed in claim 1, wherein said shared storage is positioned at local node.
4. the system as claimed in claim 1, wherein said shared storage is positioned at remote node.
5. the system as claimed in claim 1, wherein said queue comprises the store buffer district of a series of circular linkage.
6. system as claimed in claim 4, its remote node is suitable for running the process using communication channel to communicate mutually, and described communication channel selects from the group comprising TCP/IP, GPRS, WiFi, WiMax and EDGE.
7. the system as claimed in claim 1, wherein said message read/write device comprises testing fixture, for write/update, whether described testing fixture checks that described queue is whether full and be empty for reading/take out queue described in operation inspection.
8. the system as claimed in claim 1, wherein said queue comprises hereof.
9. the system as claimed in claim 1, wherein said queue is included in the file be mapped in the main memory space of processor.
10. the system as claimed in claim 1, each store buffer district wherein in queue is linked to next store buffer district and last store buffer district is linked to first store buffer district, forms the sequence of circular linkage thus to be stored in one by one in described store buffer district by all inbound messages.
11. systems as described in the arbitrary claim in all the claims, use together with wherein said system is suitable for applying with financial transaction.
Carry out the method transmitted without lock message in 12. 1 kinds of interprocess communications between at least two processes run at least one node, said method comprising the steps of:
There is provided storage facilities to store inter-process messages;
Store buffer sector is provided to arrange in shared storage in described storage facilities;
There is provided idle and point to unit, this free time is pointed to unit and points to the process context in the idle store buffer district in described queue;
There is provided at least one data to point to unit, these at least one data are pointed to unit and are pointed to the process context comprising the store buffer district of described inter-process messages;
At least one inter-process messages is received from least one process run at least one node;
Received inter-process messages are inserted the idle store buffer district pointed by described free time sensing unit of described queue;
The location updating being pointed to unit the free time is that next idle store buffer district of described queue is to hold next inter-process messages;
Take out inter-process messages from the described store buffer district comprising message pointed to pointed by unit by the corresponding data of described multiple process of described queue by multiple process simultaneously; And
Be comprise next the store buffer district by the message of the arbitrary reading by described process by the location updating that described data point to unit.
13. methods as claimed in claim 12, wherein the idle step pointing to the position of unit of renewal comprises the following steps: check whether described free time sensing unit points to the store buffer district comprising message pointed by data sensing unit, if, then wait for that this store buffer district comprising message becomes the free time, otherwise the pointer position being pointed to unit the free time is updated to next idle store buffer district.
14. methods as claimed in claim 12, the step of wherein taking out inter-process messages comprises the following steps: check that described data are pointed to unit and whether pointed to the idle store buffer district pointed to by described free time sensing unit, if so, then wait for data insertion wherein to take out.
15. methods as claimed in claim 12, the step wherein inserting the inter-process messages received comprises the following steps: by the primary memory of inserted message copy to described node.
16. methods as claimed in claim 12, the step wherein inserting the inter-process messages received comprises the following steps: perform instruction and compiler directive with predefined procedure.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| IN966/MUM/2009 | 2009-04-13 | ||
| IN966MU2009 | 2009-04-13 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101866303A CN101866303A (en) | 2010-10-20 |
| CN101866303B true CN101866303B (en) | 2015-07-01 |
Family
ID=42958036
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201010164028.0A Active CN101866303B (en) | 2009-04-13 | 2010-04-09 | High-performance message transmission system with low latency time |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN101866303B (en) |
| AU (1) | AU2010201275A1 (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102591726B (en) * | 2011-12-31 | 2014-04-09 | 青岛海信宽带多媒体技术有限公司 | Multiprocess communication method |
| CN105224258B (en) * | 2015-10-19 | 2018-05-22 | 深圳芯邦科技股份有限公司 | The multiplexing method and system of a kind of data buffer zone |
| CN105488005B (en) * | 2015-11-28 | 2018-09-07 | 贵州航天电子科技有限公司 | A kind of three cushion space communication means of 1553B buses RT sending modes |
| CN106776056B (en) * | 2016-11-10 | 2020-12-29 | 北京像素软件科技股份有限公司 | Game data processing method and device and server |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3533924B2 (en) * | 1998-01-16 | 2004-06-07 | 富士ゼロックス株式会社 | Semi-synchronous electronic conference device |
| CN101146025A (en) * | 2006-09-13 | 2008-03-19 | 华为技术有限公司 | Packet transmission method, system and compression unit for compression real time transmission protocol |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001045045A (en) * | 1999-07-29 | 2001-02-16 | Nec Corp | Selective message communication system |
| JP2007226385A (en) * | 2006-02-22 | 2007-09-06 | Fujitsu Ltd | Message queue control program and message queuing system |
-
2010
- 2010-03-31 AU AU2010201275A patent/AU2010201275A1/en not_active Abandoned
- 2010-04-09 CN CN201010164028.0A patent/CN101866303B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3533924B2 (en) * | 1998-01-16 | 2004-06-07 | 富士ゼロックス株式会社 | Semi-synchronous electronic conference device |
| CN101146025A (en) * | 2006-09-13 | 2008-03-19 | 华为技术有限公司 | Packet transmission method, system and compression unit for compression real time transmission protocol |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101866303A (en) | 2010-10-20 |
| AU2010201275A1 (en) | 2010-10-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3776162B1 (en) | Group-based data replication in multi-tenant storage systems | |
| KR102011949B1 (en) | System and method for providing and managing message queues for multinode applications in a middleware machine environment | |
| US8799213B2 (en) | Combining capture and apply in a distributed information sharing system | |
| CN100462947C (en) | Rnic-based offload of iscsi data movement function by initiator | |
| US8090926B2 (en) | Hybrid replicated shared memory | |
| US8122200B2 (en) | Modified machine architecture with advanced synchronization | |
| US8949549B2 (en) | Management of ownership control and data movement in shared-memory systems | |
| CN103257933A (en) | Method, apparatus and system for transactional memory execution in out-of-order processor | |
| US11487674B2 (en) | Virtual memory pool within a network which is accessible from multiple platforms | |
| CN101866303B (en) | High-performance message transmission system with low latency time | |
| US8392369B2 (en) | File-backed in-memory structured storage for service synchronization | |
| US20080140863A1 (en) | Multiple communication networks for multiple computers | |
| US20070100828A1 (en) | Modified machine architecture with machine redundancy | |
| CN107728935A (en) | To the data in distributed computing system again subregion | |
| US9910808B2 (en) | Reflective memory bridge for external computing nodes | |
| Mostéfaoui et al. | Time-efficient read/write register in crash-prone asynchronous message-passing systems | |
| CN102959529A (en) | Broadcast protocol for a network of caches | |
| US8140638B2 (en) | Multistage online transaction system, server, multistage online transaction processing method and program | |
| US7958322B2 (en) | Multiple machine architecture with overhead reduction | |
| US10762011B2 (en) | Reflective memory bridge for external computing nodes | |
| US11940919B2 (en) | Recall pending cache line eviction | |
| CN101647003A (en) | Providing memory consistency in an emulated processing environment | |
| Solayman et al. | Message Passing Applications: A Review | |
| Solayman et al. | Iraqi Journal of Statistical Sciences | |
| CN117785997A (en) | Data processing method and device |
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 |