CN105404546A - RDMA and HTM based distributed concurrency control method - Google Patents
RDMA and HTM based distributed concurrency control method Download PDFInfo
- Publication number
- CN105404546A CN105404546A CN201510761788.2A CN201510761788A CN105404546A CN 105404546 A CN105404546 A CN 105404546A CN 201510761788 A CN201510761788 A CN 201510761788A CN 105404546 A CN105404546 A CN 105404546A
- Authority
- CN
- China
- Prior art keywords
- data
- program
- server
- htm
- read
- 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.)
- Pending
Links
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
- G06F9/466—Transaction processing
- G06F9/467—Transactional memory
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
- G06F13/14—Handling requests for interconnection or transfer
- G06F13/20—Handling requests for interconnection or transfer for access to input/output bus
- G06F13/28—Handling requests for interconnection or transfer for access to input/output bus using burst mode transfer, e.g. direct memory access DMA, cycle steal
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides an RDMA and HTM based distributed concurrency control method. The method comprises: step 1: selecting an executive server according to data required to be accessed by a program; step 2: obtaining remote data required to be accessed by the program through the selected server and protecting the remote data; and step 3: reading all remote data, caching the remote data to the executive server, and executing the program in HTM. According to the invention, a mixed concurrency control method and the protection of the program on local and distributed data access is realized by utilizing characteristics of HTM and RDMA. For transactional support of an existing database system, the overhead of concurrency control can be reduced, the throughput of the database system can be increased, and the delay of database transaction execution can be shortened. The resource utilization rate of the database system is increased and the performance of store application services is improved, so that considerable social benefits and economic benefits can be achieved.
Description
Technical field
The present invention relates to Distributed Calculation and multinuclear computing technique field, particularly, relate to the concurrency control method for shared drive data in the calculating of distributed multi-core internal memory.
Background technology
The con current control of the concurrent internal memory calculating of distributed multi-core is the important speciality of modern computer, can facilitate the correctness of developer's estimation program, and be directly connected to the execution efficiency of application.Existing solution mainly contains the method for software, and such as two-stage lock and Optimistic Concurrency control.But these methods have larger expense under multi-core internal memory calculates and under Distributed Calculation.Two-stage lock has very large expense to the operation of each data, and Optimistic Concurrency controls the expense having the network service of many wheels under distributed environment.
Hardware transactional internal memory HTM (HardwareTransactionalMemory) is new hardware technology, and the con current control to shared drive data when directly providing executive routine by processor, has low-down expense.But hardware transactional internal memory can only use in unit.HTM provides XBEGIN to operate and allows processor enter HTM pattern, use XABORT to interrupt current execution, and again enter the program of HTM pattern after this performs after internal storage state being returned to the state before calling XBEGIN, use XEND to exit HTM execution pattern.
It is the new network communications technology that long-distance inner directly accesses RDMA (RemoteDirectMemoryAccess), is directly operated by the internal memory of network interface card to REMOTE MACHINE, has the characteristic of very high handling capacity and low delay.But the method to REMOTE MACHINE internal memory operation that RDMA provides is very limited, mainly contains and reads, write and atomic ratio replacement operation more again.This brings challenge to how utilizing RDMA operation to carry out Distributed concurrency control.
Therefore, how to utilize new hardware technology as what RDMA of HTM, rational design software agreement, reduce the expense of con current control in distributed memory calculating, promote the performance of application program and the resource utilization of system machine, become those skilled in the art's technical barrier urgently to be resolved hurrily in fact.
Summary of the invention
For defect of the prior art, the object of this invention is to provide a kind of Distributed concurrency control method based on RDMA and HTM.
According to the Distributed concurrency control method based on RDMA and HTM provided by the invention, comprise the steps:
Step 1: need the data selection of access to perform server according to program;
Step 2: the data being obtained the access of program needs by selected execution server, and data are protected;
Step 3: read total data buffer memory to after performing server, executive routine in HTM.
Preferably, described step 1 comprises: which server the data acquisition according to the access of program needs draws in the data comprising needs access, and the Transaction Information had respectively according to these servers number of entries, select to comprise the bar destination server that office will access and carry out executive routine as execution server.
Preferably, described step 2 comprises:
Step 2.1: obtain the remote data that program needs access, and remote data is protected, namely make remote data the routine access of conflict will be had to arrive by other before EOP (end of program);
Step 2.2: before program starts execution, perform server the finds program core position of remote data at other servers by distributed memory key value database, if remote data is the data needing amendment, use RDMA atomic instructions that the server state of other server at remote data place is set to locking state, otherwise establish a machine time to be used as timestamp the data of described execution server, for showing that all execution servers cannot Update Table before this timestamp;
Step 2.3: after having revised the state of all remote datas on other servers at its place, RDMA read operation is used all data to be read in the local cache performing server, wherein, the timestamp of read-only remote data is also recorded to this locality performing server.
Preferably, described step 3 comprises:
Step 3.1: after reading back at all remote datas required for program the local cache performing server, the HTM instruction that calling hardware provides enters hardware transactional memorymodel;
Step 3.2: be that local data or remote data perform different operations according to accessed data;
If the access of data is the read-write operation of local data, then, particularly,
When being read operation, check whether local data is in locking state, if be in locking state, this secondary program of interruption HTM instruction breaks that calling hardware provides performs, and rollback is carried out in the amendment of program; Otherwise carry out read operation;
When being write operation, check whether local data is in locking state, if local data stabs if having time, then judges that whether this timestamp is more late than the time of current execution server; Obtain the local data timestamp be on locking state or local data if check and be later than the time performing server, then this secondary program of interruption HTM instruction breaks that calling hardware provides performs and the amendment of rollback program; Otherwise carry out carrying out write operation on local data;
If the access of data is the read-write operation of remote data, then in the local cache performing server, carry out read-write operation at described remote data, particularly,
After program is all complete, timestamp inspection is carried out to all far-end read-only datas, if there is the timestamp of a far-end read-only data more Zao than the time of current execution server, then call the execution of this program of interruption HTM instruction breaks that hardware provides, again, after far-end read-only data being arranged and stabs large timestamp than current machine time, step 3 is performed again;
Step 3.3: after the timestamp of remote data has all checked, exits HTM transaction mode for calling hardware instruction; And use RDMA write operation that the value of remote data amendment in local cache is write back to far-end server to the remote data of program all modifications, utilize RDMA atomic operation that the state of the remote data of routine access is become init state.
Compared with prior art, the present invention has following beneficial effect:
1, the method in the present invention can interrupt the characteristic of HTM by RDMA network; in Distributed concurrency control, use HTM as main con current control; therefore the characteristic of the efficient transaction protection of HTM can be utilized; for the performance of the protection of local data access when optimizer performs, together with can being used in the guard method of remote data simultaneously.
2, the present invention utilizes the almost consistent clock of machinery compartment under distributed environment to achieve the mechanism of the lease to read-only data, shared read-only data are carried out like this by the timestamp in data, improve the concurrency that distributed program performs, under improving distributed environment further, utilize the performance of the concurrency control method of RDMA.
3, the present invention supports for existing Database Systems, can reduce the performance cost that Database Systems are brought for the strong consistency performing affairs, promotes the resource utilization of Database Systems and the performance of entirety.
Accompanying drawing explanation
By reading the detailed description done non-limiting example with reference to the following drawings, other features, objects and advantages of the present invention will become more obvious:
Fig. 1 is the principle schematic one of the Distributed concurrency control method based on RDMA and HTM provided by the invention;
Fig. 2 is the principle schematic two of the Distributed concurrency control method based on RDMA and HTM provided by the invention;
Fig. 3 be the Distributed concurrency control method calculated based on RDMA distributed memory provided by the invention the process flow diagram of remote data guard method.
Embodiment
Below in conjunction with specific embodiment, the present invention is described in detail.Following examples will contribute to those skilled in the art and understand the present invention further, but not limit the present invention in any form.It should be pointed out that to those skilled in the art, without departing from the inventive concept of the premise, some distortion and improvement can also be made.These all belong to protection scope of the present invention.
According to the Distributed concurrency control method based on RDMA and HTM provided by the invention, comprise the steps:
Step 1: need the data selection of access to perform server according to program;
Step 2: the data being obtained the access of program needs by selected execution server, and data are protected;
Step 3: read total data buffer memory to after performing server, executive routine in HTM.
Preferably, described step 1 comprises: which server the data acquisition according to the access of program needs draws in the data comprising needs access, and the Transaction Information had respectively according to these servers number of entries, select to comprise the bar destination server that office will access and carry out executive routine as execution server.
Described step 2 comprises:
Step 2.1: obtain the remote data that program needs access, and remote data is protected, namely make remote data the routine access of conflict will be had to arrive by other before EOP (end of program);
Step 2.2: before program starts execution, perform server the finds program core position of remote data at other servers by distributed memory key value database, if remote data is the data needing amendment, use RDMA atomic instructions that the server state of other server at remote data place is set to locking state, otherwise establish a machine time to be used as timestamp the data of described execution server, for showing that all execution servers cannot Update Table before this timestamp;
Step 2.3: after having revised the state of all remote datas on other servers at its place, RDMA read operation is used all data to be read in the local cache performing server, wherein, the timestamp of read-only remote data is also recorded to this locality performing server.
Described step 3 comprises:
Step 3.1: after reading back at all remote datas required for program the local cache performing server, the HTM instruction that calling hardware provides enters hardware transactional memorymodel;
Step 3.2: be that local data or remote data perform different operations according to accessed data;
If the access of data is the read-write operation of local data, then, particularly,
When being read operation, check whether local data is in locking state, if be in locking state, this secondary program of interruption HTM instruction breaks that calling hardware provides performs, and rollback is carried out in the amendment of program; Otherwise carry out read operation;
When being write operation, check whether local data is in locking state, if local data stabs if having time, then judges that whether this timestamp is more late than the time of current execution server; Obtain the local data timestamp be on locking state or local data if check and be later than the time performing server, then this secondary program of interruption HTM instruction breaks that calling hardware provides performs and the amendment of rollback program; Otherwise carry out carrying out write operation on local data;
If the access of data is the read-write operation of remote data, then in the local cache performing server, carry out read-write operation at described remote data, particularly,
After program is all complete, timestamp inspection is carried out to all far-end read-only datas, if there is the timestamp of a far-end read-only data more Zao than the time of current execution server, then call the execution of this program of interruption HTM instruction breaks that hardware provides, again, after far-end read-only data being arranged and stabs large timestamp than current machine time, step 3 is performed again;
Step 3.3: after the timestamp of remote data has all checked, exits HTM transaction mode for calling hardware instruction; And use RDMA write operation that the value of remote data amendment in local cache is write back to far-end server to the remote data of program all modifications, utilize RDMA atomic operation that the state of the remote data of routine access is become init state.
Particularly; as shown in Figure 1 and Figure 2, what the present invention designed is utilize HTM and RDMA, carrys out the logic of executive routine by HTM; and utilize RDMA to provide the protection of routine access other machines data to the method for the con current control realizing distributed memory and calculate, comprising:
Step 1: which platform machine-executed program is the data (black data in Fig. 1) that system will be accessed according to program select at.Select to have that machine the most multidata (server A in Fig. 1) will accessed when program performs to perform;
Step 2: program distally machine (server B in Fig. 1) is protected remote data; then attempt a timestamp on its state position for read-only data; then locking state is set to for the data that will revise; after protection has operated, the data (if the words of read-only data also have its timestamp) in REMOTE MACHINE are read back in the buffer memory of local machine.
Step 3a: tracking XBEGIN enters the logic that HTM carrys out executive routine.In the process that HTM performs, whenever run into will reading and writing data to far-end time, then direct to local cache read-write (as shown in Figure 2).If the read-write operation of local data, whether the state then checking data is locking state, if it is call XABORT exit HTM and re-execute, by hardware rollback program to the amendment of internal memory and the state rerun routine road logic that returns when calling XBEGIN.For write operation, further check and whether data mode has one not have the overdue time, exit HTM if also call XABORT and re-execute.If check all by; normal local data to be read and write.
Whether step 3b: the timestamp of the read-only remote data that systems inspection is all, have the timestamp of the far-end read-only data more Zao than the time of current machine, if had, calls XABORT and re-execute the guard method to far-end read-only operation in step 2.If check and pass through, then use RDMA write operation that the value of the remote data revised is write back, and change the state of these data into initialization with RDMA atomic ratio compared with replacement operation.
Further particularly, based on the method that the remote data of RDMA is protected.
In step 2, each is needed to the remote data of operation, first system can use distributed key/Value Data storehouse to obtain the core position of data, and subsequently, for the data read, system will be done as follows:
First, system can estimate a timestamp, and his value is the program runtime that the local zone time of current machine adds estimation.System initialization variable subsequently, its value is 0 (mark of the init state of Fig. 2).System is attempted the state variable of each data read, and is obtained by the core position of data, use RDMA atomic ratio more again swap operation the state atoms of data is set to this timestamp.If successfully words directly return.If failure, original value can be read back by the replacement more again of RDMA.If the value of reading back is that the time that timestamp runs machine than the program of current machine is large simultaneously, then also directly return.Otherwise, then the value of variable is become RDMA_CAS and returns the value come, repeat this process.
For the operation that will revise, system meeting initialization variable, its value is 0 (mark of the system initialization state of Fig. 2), system will be attempted the state variable of each data write, obtained by the core position of data, use RDMA atomic ratio more again swap operation the state atoms of data is set to 1 (mark of the locking state of Fig. 2) if success, directly return.If failure, original value can be read back by the replacement more again of RDMA.If rreturn value represents timestamp, simultaneously little than the time of present procedure operation machine, then variable is set to this timestamp, repeats this process; Otherwise continue this process of retry.
In sum, the Distributed concurrency control method based on RDMA and HTM that the present invention proposes, HTM can be farthest utilized to carry out the con current control of executive routine, reduce the impact brought because of software approach, RDMA network at a high speed can be utilized to reduce network overhead simultaneously, make use of the characteristic of hardware to greatest extent, the decline of the performance that the con current control reducing program brings, the performance of upper layer application can be improved greatly.
Above specific embodiments of the invention are described.It is to be appreciated that the present invention is not limited to above-mentioned particular implementation, those skilled in the art can make various distortion or amendment within the scope of the claims, and this does not affect flesh and blood of the present invention.
Claims (4)
1., based on a Distributed concurrency control method of RDMA and HTM, it is characterized in that, comprise the steps:
Step 1: need the data selection of access to perform server according to program;
Step 2: the data being obtained the access of program needs by selected execution server, and data are protected;
Step 3: read total data buffer memory to after performing server, executive routine in HTM.
2. the Distributed concurrency control method based on RDMA and HTM according to claim 1, it is characterized in that, described step 1 comprises: which server the data acquisition according to the access of program needs draws in the data comprising needs access, and the Transaction Information had respectively according to these servers number of entries, select to comprise the bar destination server that office will access and carry out executive routine as execution server.
3. the Distributed concurrency control method based on RDMA and HTM according to claim 1, it is characterized in that, described step 2 comprises:
Step 2.1: obtain the remote data that program needs access, and remote data is protected, namely make remote data the routine access of conflict will be had to arrive by other before EOP (end of program);
Step 2.2: before program starts execution, perform server the finds program core position of remote data at other servers by distributed memory key value database, if remote data is the data needing amendment, use RDMA atomic instructions that the server state of other server at remote data place is set to locking state, otherwise establish a machine time to be used as timestamp the data of described execution server, for showing that all execution servers cannot Update Table before this timestamp;
Step 2.3: after having revised the state of all remote datas on other servers at its place, RDMA read operation is used all data to be read in the local cache performing server, wherein, the timestamp of read-only remote data is also recorded to this locality performing server.
4. the Distributed concurrency control method based on RDMA and HTM according to claim 1, it is characterized in that, described step 3 comprises:
Step 3.1: after reading back at all remote datas required for program the local cache performing server, the HTM instruction that calling hardware provides enters hardware transactional memorymodel;
Step 3.2: be that local data or remote data perform different operations according to accessed data;
If the access of data is the read-write operation of local data, then, particularly,
When being read operation, check whether local data is in locking state, if be in locking state, this secondary program of interruption HTM instruction breaks that calling hardware provides performs, and rollback is carried out in the amendment of program; Otherwise carry out read operation;
When being write operation, check whether local data is in locking state, if local data stabs if having time, then judges that whether this timestamp is more late than the time of current execution server; Obtain the local data timestamp be on locking state or local data if check and be later than the time performing server, then this secondary program of interruption HTM instruction breaks that calling hardware provides performs and the amendment of rollback program; Otherwise carry out carrying out write operation on local data;
If the access of data is the read-write operation of remote data, then in the local cache performing server, carry out read-write operation at described remote data, particularly,
After program is all complete, timestamp inspection is carried out to all far-end read-only datas, if there is the timestamp of a far-end read-only data more Zao than the time of current execution server, then call the execution of this program of interruption HTM instruction breaks that hardware provides, again, after far-end read-only data being arranged and stabs large timestamp than current machine time, step 3 is performed again;
Step 3.3: after the timestamp of remote data has all checked, exits HTM transaction mode for calling hardware instruction; And use RDMA write operation that the value of remote data amendment in local cache is write back to far-end server to the remote data of program all modifications, utilize RDMA atomic operation that the state of the remote data of routine access is become init state.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510761788.2A CN105404546A (en) | 2015-11-10 | 2015-11-10 | RDMA and HTM based distributed concurrency control method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510761788.2A CN105404546A (en) | 2015-11-10 | 2015-11-10 | RDMA and HTM based distributed concurrency control method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105404546A true CN105404546A (en) | 2016-03-16 |
Family
ID=55470046
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510761788.2A Pending CN105404546A (en) | 2015-11-10 | 2015-11-10 | RDMA and HTM based distributed concurrency control method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105404546A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106844048A (en) * | 2017-01-13 | 2017-06-13 | 上海交通大学 | Distributed shared memory method and system based on ardware feature |
| CN107888657A (en) * | 2017-10-11 | 2018-04-06 | 上海交通大学 | Low latency distributed memory system |
| CN110069431A (en) * | 2018-01-24 | 2019-07-30 | 上海交通大学 | Elastic Key-Value key-value pair data storage method based on RDMA and HTM |
| CN113238856A (en) * | 2021-03-09 | 2021-08-10 | 西安奥卡云数据科技有限公司 | RDMA (remote direct memory Access) -based memory management method and device |
| CN113448897A (en) * | 2021-07-12 | 2021-09-28 | 上海交通大学 | Array structure and optimization method suitable for pure user mode remote direct memory access |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5224215A (en) * | 1990-07-13 | 1993-06-29 | International Business Machines Corporation | Message queue processing among cooperative processors having significant speed differences |
| US20020178210A1 (en) * | 2001-03-31 | 2002-11-28 | Manoj Khare | Mechanism for handling explicit writeback in a cache coherent multi-node architecture |
| US6856619B1 (en) * | 2000-03-07 | 2005-02-15 | Sun Microsystems, Inc. | Computer network controller |
| US20060236063A1 (en) * | 2005-03-30 | 2006-10-19 | Neteffect, Inc. | RDMA enabled I/O adapter performing efficient memory management |
| CN102012933A (en) * | 2010-12-02 | 2011-04-13 | 清华大学 | Distributed file system and method for storing data and providing services by utilizing same |
| CN103618804A (en) * | 2013-12-16 | 2014-03-05 | 北京航空航天大学 | Performance difference-based load balancing method for distributed key value storage system |
| US20150317182A1 (en) * | 2014-05-05 | 2015-11-05 | Google Inc. | Thread waiting in a multithreaded processor architecture |
-
2015
- 2015-11-10 CN CN201510761788.2A patent/CN105404546A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5224215A (en) * | 1990-07-13 | 1993-06-29 | International Business Machines Corporation | Message queue processing among cooperative processors having significant speed differences |
| US6856619B1 (en) * | 2000-03-07 | 2005-02-15 | Sun Microsystems, Inc. | Computer network controller |
| US20020178210A1 (en) * | 2001-03-31 | 2002-11-28 | Manoj Khare | Mechanism for handling explicit writeback in a cache coherent multi-node architecture |
| US20060236063A1 (en) * | 2005-03-30 | 2006-10-19 | Neteffect, Inc. | RDMA enabled I/O adapter performing efficient memory management |
| CN102012933A (en) * | 2010-12-02 | 2011-04-13 | 清华大学 | Distributed file system and method for storing data and providing services by utilizing same |
| CN103618804A (en) * | 2013-12-16 | 2014-03-05 | 北京航空航天大学 | Performance difference-based load balancing method for distributed key value storage system |
| US20150317182A1 (en) * | 2014-05-05 | 2015-11-05 | Google Inc. | Thread waiting in a multithreaded processor architecture |
Non-Patent Citations (2)
| Title |
|---|
| XINGDA WEI等: "Fast In-memory Transaction Processing using RDMA and HTM", 《SOSP "15》 * |
| 董勇，等: "基于Infiniband技术构建高性能分布式文件系统-Lustre", 《计算机工程与应用》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106844048A (en) * | 2017-01-13 | 2017-06-13 | 上海交通大学 | Distributed shared memory method and system based on ardware feature |
| CN106844048B (en) * | 2017-01-13 | 2020-11-06 | 上海交通大学 | Distributed memory sharing method and system based on hardware characteristics |
| CN107888657A (en) * | 2017-10-11 | 2018-04-06 | 上海交通大学 | Low latency distributed memory system |
| CN110069431A (en) * | 2018-01-24 | 2019-07-30 | 上海交通大学 | Elastic Key-Value key-value pair data storage method based on RDMA and HTM |
| CN110069431B (en) * | 2018-01-24 | 2020-11-24 | 上海交通大学 | Elastic Key-Value Key Value pair data storage method based on RDMA and HTM |
| CN113238856A (en) * | 2021-03-09 | 2021-08-10 | 西安奥卡云数据科技有限公司 | RDMA (remote direct memory Access) -based memory management method and device |
| CN113238856B (en) * | 2021-03-09 | 2022-07-26 | 西安奥卡云数据科技有限公司 | RDMA-based memory management method and device |
| CN113448897A (en) * | 2021-07-12 | 2021-09-28 | 上海交通大学 | Array structure and optimization method suitable for pure user mode remote direct memory access |
| CN113448897B (en) * | 2021-07-12 | 2022-09-06 | 上海交通大学 | Optimization method suitable for pure user mode far-end direct memory access |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107491345B (en) | Method for writing picture data and distributed NewSQ L database system | |
| US10936578B2 (en) | Client-driven commit of distributed write transactions in a database environment | |
| CN105404546A (en) | RDMA and HTM based distributed concurrency control method | |
| US8364909B2 (en) | Determining a conflict in accessing shared resources using a reduced number of cycles | |
| US10572510B2 (en) | Distributed database transaction protocol | |
| CN108363806B (en) | Multi-version concurrency control method and device for database, server and storage medium | |
| US8996811B2 (en) | Scheduler, multi-core processor system, and scheduling method | |
| KR101970390B1 (en) | Lock elision with binary translation based processors | |
| US20110167222A1 (en) | Unbounded transactional memory system and method | |
| US9471400B1 (en) | Reentrant read-write lock algorithm | |
| EP2797014A1 (en) | Database update execution according to power management schemes | |
| CN104598397A (en) | Mechanisms To Accelerate Transactions Using Buffered Stores | |
| CN113127212B (en) | Method and device for processing transaction in blockchain | |
| JP2007501455A (en) | Parallel recovery with non-failing nodes | |
| US9652169B2 (en) | Adaptive concurrency control using hardware transactional memory and locking mechanism | |
| CN112527735A (en) | Data merging method and device applied to key value storage system | |
| US20180300147A1 (en) | Database Operating Method and Apparatus | |
| CN104462497A (en) | Crossing-base transaction processing method and system based on relational databases | |
| CN111581234A (en) | RAC multi-node database query method, device and system | |
| US11609752B2 (en) | Loop lock reservation | |
| US20200050601A1 (en) | Hardware transactional memory (htm) assisted database transactions | |
| US20120059997A1 (en) | Apparatus and method for detecting data race | |
| CN110520845A (en) | Update HTM user's abort metadata | |
| CN106959989A (en) | The synchronous method and device in master/slave data storehouse | |
| US20180300132A1 (en) | Method and system for restructuring of collections for synchronization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160316 |
|
| RJ01 | Rejection of invention patent application after publication |