CN1190741C - Apparatus and method for supporting data transactions - Google Patents

Apparatus and method for supporting data transactions Download PDF

Info

Publication number
CN1190741C
CN1190741C CNB011258179A CN01125817A CN1190741C CN 1190741 C CN1190741 C CN 1190741C CN B011258179 A CNB011258179 A CN B011258179A CN 01125817 A CN01125817 A CN 01125817A CN 1190741 C CN1190741 C CN 1190741C
Authority
CN
China
Prior art keywords
data
memory devices
tsld
affairs
block
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.)
Expired - Fee Related
Application number
CNB011258179A
Other languages
Chinese (zh)
Other versions
CN1365056A (en
Inventor
J·萨特兰
I·戈德
D·舍恩瓦德
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
International Business Machines Corp
Original Assignee
International Business Machines Corp
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Priority claimed from US09/651,910 external-priority patent/US6668304B1/en
Application filed by International Business Machines Corp filed Critical International Business Machines Corp
Publication of CN1365056A publication Critical patent/CN1365056A/en
Application granted granted Critical
Publication of CN1190741C publication Critical patent/CN1190741C/en
Anticipated expiration legal-status Critical
Expired - Fee Related legal-status Critical Current

Links

Images

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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
    • G06F11/14—Error detection or correction of the data by redundancy in operation
    • G06F11/1402—Saving, restoring, recovering or retrying
    • G06F11/1474—Saving, restoring, recovering or retrying in transactions

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Quality & Reliability (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Signal Processing For Digital Recording And Reproducing (AREA)

Abstract

Apparatus for performing data transactions, comprising at least one memory device for receiving data as successive data blocks for storage at respective physical locations therein, and control circuitry. The control circuitry is configured to transfer the contiguous data blocks, at least some of the data blocks, and a parameter indicative of an identity of the data transaction, an identity of the transaction status recorder, and a location of a subsequent data block in the contiguous data blocks to the at least one memory device when writing to the at least one memory device. The control circuitry is also configured to transition the data state to complete (commit or terminate) the data transaction using some of the indicated parameters. In the event of a device failure before a transaction completes (commits or terminates), the device causes recovery to be completed with the data stored in the data block.

Description

The equipment and the method that are used for support data affairs
Technical field
The present invention relates generally to data transactions, specifically is the distributed transaction support that data write volatile memory.
Background technology
Affairs are meant a series of message exchanges and the related work (for example database update) of the separability unit that is considered as guaranteeing data integrity.In affairs, data are transformed into another consistent state from a consistent state.In order to finish affairs and to make data variation become permanent (or " submitted "), about affairs failure aspect, must guarantee that affairs are atoms, promptly these affairs must intactly be finished.If certain situation occurred before these affairs are completed successfully, then these affairs are terminated, and must cancel the change that data are made, so that produce such result, promptly these affairs never existed.Hereinafter, when term " being finished " when being applied to affairs, be meant these affairs or submitted or be terminated.
When the participator in the affairs was a distributed system a part of, the intrinsic difficulty of affairs support just was exacerbated.Thereby be necessary to guarantee that affairs atomically and are as one man submitted to by all participators or stop.For example, some participator in affairs may fail, and in a distributed system, certain among possible other participator is not known this failure.Equally, the participator who recovers from failure must determine the destiny of these affairs.
Be hereby incorporated by, people such as Bernstein (Addison-Wesley, in the 7th chapter of 1987) " parallel control in the Database Systems and recovery ", the consistent affairs atom committing protocol (ACP) that is used for guaranteeing on a plurality of nodes of distributed system has been described.The author has described one two sections submissions (2PC) agreement as the ACP example.Whether the 2PC agreement comprises the phase one, wherein should submitted or termination to all these affairs of participator's poll of affairs.In the subordinate phase of 2PC agreement, a coordinator of these affairs determines that based on poll these affairs are whether submitted or stops, and this is determined to send to the participator.
Known in the art in a computer system event of failure, the method for effectively storing data and from stored data, recovering.These methods depend on the nonvolatile memory of the information stores to outside the data, disk typically, and when breaking down, use this extra information to recover stored data.
United States Patent (USP) 5,345,575 this people such as English that its disclosure is incorporated herein by reference has been described the Magnetic Disk Controller that comprises a storer.This storer contains the logical address that will be stored in the data block in the disk and is mapped to a form that has identified physical memory cell.In writing data into a storage unit, Magnetic Disk Controller also writes relevant logical address, the time mark of each storage unit and has indicated a certain data block to appear at the data where of a series of data blocks.By reading from whole magnetic disk in fact, these extra information are used to recover from the system failure.
United States Patent (USP) 5,481,694 this people such as Chao that its disclosure is incorporated herein by reference has been described the electronic data memory system that comprises a storer, a plurality of disk drive and a controller.The form of the physical address on this storer has comprised cross reference logical address and the disk drive, the list of physical addresses that contains the list of physical addresses of stale data and can receive the section on the disk drive of data.When data are write disk drive, comprise for the logical address of polylith and the mark of sequence number being write with data.In order from the system failure, to recover, be stored in checkpoint record and check point section recovery form and tabulation on the disk drive.
In the 15-28 page or leaf with the procceedings of the 14th the Principles of Operating System meeting in Dec, 1993, being entitled as of people such as de Jonge, " (logic magnetic disc: a new method improving file system) " article was hereby incorporated by The Logical Disk:A New Approach toImproving File System, wherein the author has described a kind of logic magnetic disc, has wherein stipulated to be used for the interface of magnetic disk memory that file management is separated with disk management.This interface uses logical block number (LBN) and piece tabulation, and supports multifile system.The author claims and supports an atom recovery unit (ARU).In rejuvenation, all logic magnetic disc orders that belong to identical ARU are counted as single invisible operation.Therefore, the state that exists before logic magnetic disc always returns to, or return to the state that exists after all operations of having carried out ARU.But, do not support parallel ARU.
" (Loge: a kind of Magnetic Disk Controller of self) " article is hereby incorporated by Loge:a self-organizing diskcontroller, and wherein the author has described and used a conversion table and distribution diagram with the system of data storage in the disk with being entitled as of the people such as English in the 237-251 page or leaf of the USENIX winter 1992 technical conference procceedingss in January, 1992.The trailer label that will comprise a block address and a time mark writes disk with stored data.Information in the trailer label system that makes can recover from fault.
Be entitled as " Mime:a high performance parallelstorage device with strong recovery guarantees (Mime: have the high performance parallel memory devices that powerful recovery guarantees) " with people such as Chao, the article of HPL-CSP-92-9 (Hewlett-packard company published in November, 1992) is hereby incorporated by, and wherein the author has described the disk storage structure that is similar to Loge described above.In Mime, trailer label comprises the mark of a block address writing for polylith, sequence number and the polylith last grouping in writing.With the same among the Loge, the trailer label information system that makes can recover from fault.
Mime supports the atom polylith of the finite form affairs support in the visibility group form to write.Mime guarantees under the situation of fault, and all pieces in activity visibility group are write and are terminated.
Summary of the invention
A purpose of some aspect of the present invention provides the improvement system of support data affairs of being used for.
Another purpose of some aspect of this aspect provides when the participator of affairs is distributed on the network, is used for carrying out the improvement system of parallel data affairs.
In the preferred embodiment of this aspect, one or more memory devices are fixed disk preferably, is used to the data content of the affairs that one or more client started of memory storage devices.In these memory devices each is managed by the control circuit of a storage server preferably, in the selected frame of this server with the data content write storage device of affairs.This memory device is called affairs here and supports logic magnetic disc (TSLD).Storage server has volatile memory, and data structure is wherein arranged, and the value of data structure is used to write down the Transaction Information that is written to TSLD, and the physics and the logical address of the piece frame that is written into of dynamic link data.
In order to protect the fault of a particular memory server of a TSLD defence, the service test point operation stores the value in the data in server structure among the TSLD into all period interval.Between the check point operation, the value in the data structure is also stored by the data content of the affairs in the piece frame of each TSLD.Value in the data structure and check point data preferably make the piece frame to be found easily in the incident of storage server fault, so that the data content in the piece frame can be resumed.In event of failure, storage server reads the also process of " playback " TSLD operation of stored check point data, has comprised the submission since carrying out check point last time and stops affairs.Playback procedure makes storage server to recover the state of its state and any ongoing affairs when fault.Above-mentioned temporary patent application and transfer the same assignee's of the application, another that submitted on July 6th, 2000 is entitled as " Enhanc ed Stable Disk Storage's (enhancing fixed disk storage) ", the alternative method of having described storage and recover in this patented claim that its disclosure is incorporated herein by reference.
The preferred embodiment of this aspect is preferably supported as one two sections submissions (2PC) agreement described in the background of the present invention.In the phase one of affairs, the person's that is called the transaction coordination specific T SLD client is appointed as main TSLD by one among the TSLD that will participate in these affairs and starts this affairs.This main TSLD works as the register that writes down transaction status, and the client instructs this main TSLD to begin this affairs.The management server of other TSLD of client by participating in these affairs instructs appropriate data content " soft write " separately the TSLD of these TSLD with these affairs, thereby produces the piece value of not submitting to, and notifies client under every kind of situation to operate to be done.In a soft write operation, logic that the piece frame is written into and physical address are linked temporarily.In case client's phase one is completed successfully, then in subordinate phase, it just instructs main TSLD to submit this affairs to, and the client issues each TSLD of submitting to instruction to give to participate in submitting this affairs to then, thereby produces submitted piece value.Carry out submit operation by the address that permalink is linked temporarily.If in a preset time section, the phase one is not completed successfully, then the client instructs main TSLD to stop this affairs, and the client issues each TSLD that command for stopping give to participate in stopping this affairs then, thereby the address that exists before these affairs link is resumed.
For the transaction operation on the transaction system known in the art, there are a plurality of significant advantage in above-described system.Especially:
Identical information in the data structure is used to make a TSLD to recover fully basically from server failure, and allows effective execution of two-phase commitment protocol.
System supports the distributed transaction on a plurality of TSLD, also supports the parallel data affairs on one or more TSLD.
There is not the central daily record of transaction status.Each affairs can be selected different main TSLD, thereby the transaction status record is distributed.
TSLD supports the concurrent activities affairs and can recover from a plurality of transaction failures.
Because all information in the data structure all are stored among the TSLD, so TSLD can move to a workspace server from a failed server, thereby the address that can regenerate open affairs links and information.
Do not need extra I/O operation, just can carry out storage the increase of the data structure information of TSLD to TSLD.
Even when system comprises respectively a plurality of disks on a network and/or a plurality of server, soft write operation is with the submission of affairs and stop also being supported.
Can realize the check point operation with flexible way, for example as a consistency operation or by identifying the composition in the data structure that behind last check point, changes.
The recovery of TSLD after the fault is fast, owing to have only the stored frame in this TSLD check point last time operation back to be read.
The recovery of TSLD after the fault produces about the enough information of open affairs so that the participator of these affairs can the inquiry transaction state and obtained whether to submit to or stop the consensus of these affairs.Under the situation of client's fault, there is enough information to be used for submitting to or stopping affairs on the TSLD, its information can be influenced by other client.
Therefore,, provide the equipment of support data affairs of being used for, having comprised according to the preferred embodiments of the invention:
Be used for receiving data as consecutive data block to be stored at least one TSLD of physical location separately; And
Control circuit, be configured to when writing at least one TSLD, with consecutive data block, at least some data blocks and indicated in the consecutive data block at least one parameter of the physical location of data block afterwards to be sent at least one TSLD together, and be configured to the state of translation data, so that utilize at least one parameter to finish data transactions.
Control circuit preferably includes at least one communication link that control circuit is coupled at least one TSLD, and control circuit is by at least one TSLD of at least one link management.
The said equipment preferably includes a data transmission network, is used for data being sent to control circuit and being coupled with this control circuit.
Data transmission network preferably is coupled to and transfers data at least one TSLD in addition, and control circuit is managed at least one TSLD by this data transmission network.
The said equipment preferably includes a client who data is sent to control circuit.
Control circuit preferably includes at least one volatile memory, be used for storing one or more data structures of the data that contain one or more characteristics of having indicated at least some data blocks, in these data at least some are written at least one TSLD by control circuit, and the content of at least one volatile memory can be regenerated at least some data in the data in the one or more data structures from be stored at least one TSLD like this.
In addition, one in the data structure preferably includes a conversion table, is used for physical address and in the mark data block which that LBA (Logical Block Addressing) with consecutive data block is mapped to separately to participate in the data transactions of an opening.
One in the data structure preferably includes an auxiliary conversion table, be used for being each open data transactions separately, the LBA (Logical Block Addressing) relevant with open affairs is mapped at the physical address separately that starts the consecutive data block that exists before these affairs, and is mapped to a sign of open data transactions.
Control circuit preferably is written to continuous data block with logical address separately.
Control circuit preferably is written to continuous data block with the sign of data affairs.
Control circuit preferably is written to continuous data block with the sign of the register of the state of data affairs.
One in the data structure preferably includes an assignment bit map, is used for shining upon each the availability in the consecutive data block.
In addition, a pointer value that preferably includes the unit of a next available block in the consecutive data block in the data structure.
A pointer value that preferably includes the unit of first available block in the consecutive data block in the data structure.
At least one TSLD preferably includes a disk that magnetic head is arranged, and control circuit is on the surface at disk on the single direction, and a series of scannings with magnetic head are written to data block in the disk.
In addition, each in a series of scannings preferably has a check period, and in the data structure one comprises a value of having indicated current check period.
In the data structure one preferably includes a work allocation bitmap copy, is used for being stored in each the availability in the consecutive data block at one or more ends of scan place of magnetic head.
When one or more scannings of magnetic head were finished, control circuit preferably write at least some data in one or more data structures among at least one TSLD.
One in the data structure preferably includes a transaction status daily record, be used for storing at least one record of data transactions, wherein at least one record comprises a sign of data transactions and the sign of participating at least one TSLD of this data transactions, wherein at least one record comprises at least one in the group record, this group record comprises: beginning-affairs, submission-affairs and stop affairs, they write down respectively affairs begin, submitted and be terminated.
Control circuit is preferably in and writes afterbody on the continuous data block, and this afterbody comprises one or more data fields of having indicated each data block to use.
One or more data fields preferably include one of data transactions sign and as the sign of the main TSLD of data transactions state recording device.
One or more data fields preferably include a logical address of data block.
One or more data fields preferably include a mark of indicating the data transactions state.
This equipment preferably includes a Magnetic Disk Controller, and it controls at least one TSLD, and wherein Magnetic Disk Controller can write the content of recovering a volatile memory the data of at least one TSLD from control circuit.
At least one parameter preferably includes a plurality of parameters.
Data transactions preferably includes a plurality of parallel data affairs.
At least one TSLD preferably includes a disk by server admin.
At least one TSLD preferably includes a plurality of TSLD.
At least one parameter preferably includes a transaction status daily record, and it has write down one or more states of data transactions.
At least one parameter preferably includes the part of at least one distributed transaction state daily record, this log record one or more states of data transactions.
At least one memory device preferably is used as transportable, so that operated by other control circuit, thereby recovery is stored in the information at least one memory device.
According to the preferred embodiments of the invention, a kind of method that is used for carrying out data transactions also is provided, comprising:
In at least one TSLD, receive each of data as consecutive data block, at least some data blocks and indicated the parameter of the physical location of a subsequent data piece in the consecutive data block; And
In response at least one parameter, change the state of the data among at least one TSLD, so that finish data transactions.
At least one parameter preferably includes a plurality of parameters.
The reception data preferably include data are sent at least one TSLD by data transmission network at least one TSLD.
The transmission data preferably include at least one TSLD are connected to data transmission network.
The transmission data preferably include:
By control circuit at least one TSLD is connected to data transmission network; And
Control circuit is connected to data transmission network.
In addition, the reception data preferably include from a client and receive data.
This method preferably includes one or more data structure storage at least one volatile memory, above-mentioned data structure contains the data of one or more characteristics of having indicated at least some data blocks, and at least some data in the data structure are written among at least one TSLD, so that recover in the content of at least one read-only storage at least some data in can the one or more data structures from be stored at least one TSLD.
Store one or more data structures and preferably include conversion table of storage, this conversion table is used for LBA (Logical Block Addressing) with consecutive data block and is mapped to separately physical address, and in the mark data block which is current is participating in an open data transactions.
Store one or more data structures and preferably include auxiliary conversion table of storage, be used for being each open data transactions separately, and the LBA (Logical Block Addressing) relevant with open affairs is mapped to the physical address separately of the consecutive data block that existed before the log-on data affairs, and be mapped to the sign of this opening data transactions.
Receiving data preferably includes separately logical address is write continuous data block.
This method preferably includes uses conversion table to locate a certain data block, so as from certain data block reading of data.
Store one or more data structures and preferably include position bitmap of having indicated the availability of each in the continuous position of storage.
In addition, at least some data being write at least one TSLD preferably includes and uses the following step that data are write in the consecutive data block one:
Scan one or more data structures to determine the physical location of an available block among at least one TSLD;
Data and at least some contents of one or more data structures are write physical location; And
In response to the physical location that is determined, upgrade one or more data structures.
Scanning one or more data structures preferably includes and checks whether a logical address is included in the open affairs.
Write step that the data of one or more data structures and at least some contents comprise preferably:
Write value corresponding to a soft write operation; And
Write the sign as at least one TSLD of main TSLD, to carry out interim write operation, this TSDL is as a register job of data transactions.
The data that write one or more data structures preferably include with at least some contents and write one corresponding to the value that begins to operate, and upgrading one or more data structures comprises and writes an opening entry, this record comprises a sign of affairs and the sign of main TSLD, begins operation so that carry out.
Write the data of one or more data structures and at least some contents and preferably include the value that writes corresponding to a submit operation, and upgrade one or more data structures and comprise:
The mark of the open state of affairs has been indicated in removing;
Write one and submit record to, this record comprises a sign of affairs and the sign of main TSLD, so that carry out submit operation; And
Upgrade a transaction status daily record.
Write the data of one or more data structures and at least some contents and preferably include the value that writes corresponding to a terminating operation, and upgrade one or more data structures and comprise:
The mark of the open state of affairs has been indicated in removing;
Write one and stop record, this record comprises a sign of affairs and the sign of main TSLD, so that the executive termination operation; And
Upgrade a transaction status daily record.
This method preferably includes the state of the one or more data structures of scanning with the specified data affairs.
This method preferably includes carries out a check point operation that comprises the following steps:
Pin one or more data structures;
The content of one or more data structures is written to the check point position of at least one TSLD; And
In response to content being write at least one TSLD, change at least some contents of one or more data structures.
This method preferably includes carries out the recovery operation that comprises the following steps:
From at least one TSLD, read the content of one or more data structures; And
In response to foregoing, upgrade the one or more data structures at least one volatile memory.
Carry out recovery operation and preferably include the content that reads since carrying out all one or more data structures that have been written into since the check point operation.
In addition, carry out recovery operation and preferably include the content that reads since carrying out all one or more data structures that write since the check point operation, this reads the used time and no more thanly basically is used for writing all used time of one or more data structures that has write since the check point operation since carrying out.
Carry out recovery operation preferably include the state of determining open affairs and the one or more clients that make open affairs can query State and determine whether submission or stop this opening affairs.
This method preferably includes carries out a transaction coordination operation that comprises the following steps:
Select a register among at least one TSLD as the data transactions state;
Submit to a beginning transaction operation to give register;
Be that a schedule time of coming self-recorder is waited in a response, the state of data transactions has been indicated in this response;
Based on above-mentioned response, submit to one or more soft write operations to give at least one TSLD; And
Based on above-mentioned response, determine whether to submit to or stop this affairs, submit or stop transaction operation to by sending one.In addition, this method preferably includes and carries out the affairs complete operation that comprises the following steps:
From at least one TSLD that contains the content of having indicated a data transaction status, read first group of consecutive data block; And
In response to the content of first group of consecutive data block, second group of consecutive data block write among at least one TSLD, and one group of one or more data structure is write at least one volatile memory.
Data transactions preferably includes a plurality of parallel data affairs.
At least one TSLD preferably includes a disk by server admin.
At least one TSLD preferably includes a plurality of TSLD.
According to a preferred embodiment of the invention, also provide the equipment that is used for electronic data storage, having comprised:
A TSLD is used for receiving consecutive data block to store physical location separately into; And
Control circuit is configured when writing TSLD, and consecutive data block is sent to TSLD with at least some data blocks, one or more parameter of the physical location of the data block afterwards in the consecutive data block of having indicated.
One or more parameters preferably include a plurality of parameters.
According to embodiment of the present invention, the method that is used for electronic data storage also is provided, comprising:
Provide consecutive data block to store among the TSLD physical location separately into;
For each of at least some data blocks in the consecutive data block has determined to indicate one or more parameters of the physical location of a data block afterwards in this consecutive data block; And
Consecutive data block and one or more parameter are stored among the TSLD.
One or more parameters preferably include a plurality of parameters.
According to a preferred embodiment of the invention, a computer software product that is used for carrying out data transactions also is provided, comprise programmed instruction is recorded in wherein computer-readable medium, when instruction is wherein read by computing machine, make computing machine at least one TSLD, receive data as consecutive data block, in at least some data blocks each with indicated in the consecutive data block one at least one parameter of the physical location of data block afterwards, and in response at least one parameter, change the state of the data among at least one TSLD, to finish data transactions.
According to a preferred embodiment of the invention, a computer software product that is used for electronic data storage also is provided, the computer-readable medium that comprises the instruction that wherein has program recorded thereon, when instruction is wherein read by computing machine, make computing machine provide consecutive data block to store among the TSLD physical location separately into, think that at least some data blocks in the consecutive data block each determined to indicate one or more parameters of the data block physical location afterwards in the consecutive data block, and consecutive data block and one or more parameter are stored among the TSLD.
Now understand the present invention more fully according to the description and the accompanying drawing of the following preferred embodiment of the invention, wherein:
Description of drawings
Figure 1A is according to a preferred embodiment of the invention, has provided the schematic block diagram of a distributed memory system, and wherein disk is arrived storage server and wherein affairs generation by physical connection;
Figure 1B is according to a preferred embodiment of the invention, provided the schematic block diagram of the distributed memory system that substitutes, wherein disk by virtual be connected to storage server and wherein affairs take place;
Fig. 2 is according to a preferred embodiment of the invention, has provided the affairs of support logic magnetic disc (TSLD) of the system that is included among Figure 1A or Figure 1B and the schematic block diagram of the relation between the server;
Fig. 3 is according to a preferred embodiment of the invention, by block structured schematic block diagram on the disk of TSLD storage;
Fig. 4 is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of a step in the soft write operation;
Fig. 5 is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of a step in the soft read operation;
Fig. 6 is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of a step in the read operation;
Fig. 7 is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of a step in beginning to operate;
Fig. 8 is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of a step in the submit operation;
Fig. 9 is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of a step in the terminating operation;
Figure 10 A is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of a step in the query manipulation;
Figure 10 B is according to a preferred embodiment of the invention, from the constitutional diagram of the possible outcome of the query manipulation of Figure 10 A;
Figure 11 is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of a step in the check point operation;
Figure 12 is according to a preferred embodiment of the invention, and expression is included in a process flow diagram for the step in the coordinated manipulation of affairs;
Figure 13 is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of the step in first recovery operation;
Figure 14 is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of the step in second recovery operation;
Figure 15 is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of the step in the 3rd recovery operation;
Figure 16 is according to a preferred embodiment of the invention, and expression is included in the process flow diagram of the step in the affairs complete operation.
Embodiment
Figure 1A is discussed now, and it is according to a preferred embodiment of the invention, has represented a wherein schematic block diagram of the distributed memory system 20 of affairs generation.System 20 comprises a plurality of disks 28,30 and 34 as the memory device operation.Disk 34 is connected to storage server 32, comprises the control circuit 33 that makes that server 32 can hyperdisk.Disk 28 and 30 is connected to storage server 26, comprises the control circuit 35 and 37 that makes that respectively server 26 can hyperdisk.By the memory device of appropriate control circuit management, for example disk 28,30 and 34 is called affairs support logic magnetic disc (TSLD) here.Here the control circuit that is called the TSLD interface is carried out memory device or is connected to TSLD operation on the equipment on the circuit.Provided the detailed description of TSLD, TSLD interface and TSLD operation below.Each disk 28,30 and 34 in the system 20 is assigned with a unique sign, is called TSLD here respectively 1, TSLD 2, TSLD 3Although system 20 comprises three TSLD, the system of being to be understood that can comprise the TSLD of any amount, and wherein each connects a storage server (the following server that is called simply) and managed by it.It is also understood that the one or more TSLD of each server admin in the system 20.
Server 26 and 32 and then be connected to a data transmission network 24 again, LAN (Local Area Network) for example, client 22 is connected to this network.Client 22 knows the sign of each disk in the system 20 and knows the server separately of each disk of management.Client 22 coordinates the execution of TSLD affairs.It starts a TSLD operation, and this TSLD operation is forwarded to appropriate storage server or server with the sign of operating one or more TSLD that will be applied to.In case receive a TSLD operation, each server is carried out by the operation on one or more TSLD of client's 22 appointments, and the client is made response.
Figure 1B is according to a preferred embodiment of the invention, the schematic block diagram of the distributed memory system 40 that the wherein affairs that expression substitutes take place.Except the difference that describes below, the operation of system 40 generally is similar to the operation (Figure 1A) of system 20, and wherein the element by the indication of the identical reference numerals in the system 40 and 20 all is identical in structure with operating generally.System 40 comprises server 26 and 32 and disk TSLD 1, TSLD 2, TSLD 3A storage area network 42 that is connected to.Server 26 and 32 is connected to and manages its disk separately by network 42 by virtual, shown in the dotted line among the figure.Transmit a TSLD by network 42 and operate to appropriate server or a plurality of server, client 22 starts this operation, and each particular server is by network customer in response 22.
Fig. 2 is according to a preferred embodiment of the invention, the schematic block diagram of the relation between expression server 26 and the TSLD28.Although following description is applied to be included in TSLD 28 and relevant TSLD interface 35 thereof in the server 26, be to be understood that each TSLD in system 20 and/or the system 40 and the description of separately following basically TSLD of regarding to 28 of server and server 26 come mutual.
TSLD 28 comprises an on-fixed memory disk 62 by disc driver 61 operations.Disk 62 realizes in magnetic medium 63 that preferably it is by 78 read-writes of the magnetic disk head on the disk support 79.As described in more detail below, to write at piece, disk support 79 is with " mobile forward direction storage " motion action.The operation of Magnetic Disk Controller 76 control disks 62.Magnetic Disk Controller 76, disc driver 61 and all parts thereof preferably popular, the industry standard item.Utilize methods known in the art, magnetic medium 63 is divided into a plurality of physical sector 64a, 64b, 64c ... 64z is referred to as sector 64 here.Each physical sector is pointed to by a physical sector address.The size of each sector preferably equals the byte of 2 integral number power, for example 512 bytes.
Be stored in piece frame position 82a, 82b, 82c that data on the disk 62 are written into a plurality of fixed measures ..., general designation is made piece frame 82 here.Each piece frame is preferably pointed to by the physical address sector of first sector, and this first sector is made up of the piece frame.Each piece frame 82 preferably is made of an integer contiguous sector 64, and each piece frame can be stored data block.In the formatting procedure of disk 62, some sectors 64 are preserved for describing physics and the logic parameter that relates to server 26 and disk 62.Parameter comprises the size of each sector 64, the quantity of the sector in the disk 62, the scope of the LBA (Logical Block Addressing) (LBA) that the size of piece frame 82 and quantity and disk 62 are supported.What keep in the formatting procedure of disk 62 equally is the space of using in the check point operating process, and wherein except other affairs, the data of storage are stored in the disk 62 in the volatile memory 70 that comprises in the server 26.Be described in more detail below the check point operation.
To describe a plurality of fixedly dynamic data structures of volatile memory 70 storages below in more detail, these data structures are as 35 work of a TSLD interface.The content of data structure is preferably by server 26 controls.
Storer 70 is stored as interface 35:
A conversion table data structure 72, it is changed between LBA and physical block address.This conversion table is called TT (i) here with the respective physical address binding of each LBA (Logical Block Addressing) " i " with a piece frame of disk 62.TT (i) preferably holds the content of nearest stored logical address for the piece of " i ".At first, all values of TT (i) all is configured to NULL (sky).When data are written into disk 62, just change for the value of the TT (i) of certain logic block address i, and be updated to disk sector address into stored block space piece i last time from the NULL value.Conversion table 72 also comprises a tag field 73, and the mark of each TT (i) in the conversion table 72 is provided to that indicator dog " i " is current is participating in open affairs.When a specific markers was set up, item TT (i) separately pointed to a piece frame of not submitting to.Describe as following, it is found to use an auxiliary transformation table entries ATT (i) to come for the submitted piece frame of piece i.If TT (i) Xiang Wei is labeled, then TT (i) points to the submitted piece frame for piece i.
An auxiliary conversion table (ATT) data structure 172 is the forms of simplifying of table 72, comprises the nonempty entry of table 72, and wherein tag field 73 is set up.In other words, auxiliary conversion table 172 writes down the current piece in the open affairs that is included in.Table 172 comprises for the item that is included in each the piece i in the open affairs, the corresponding piece frame address ATT (i) of stored i before open affairs, and sign ' the Trans ID ' of the current affairs that are comprised in of piece i.
An assignment bit map data structure 74 is used at piece locating available disk block frame 82 in each storing process of disk 62.For each piece frame i, the if block frame is available for the storage of piece content, and then the corresponding bits in the assignment bit map 74 is configured to 0.The if block frame contains the content of a piece, and perhaps the if block frame has been retained by server 26 data structures and has used, and then corresponding bit is configured to 1.When disk 62 is initialised, make except those all piece frames that kept for server 26 all become available so that each bit in the assignment bit map 74 all is configured to 0.
A work allocation bitmap data structure 174.Working position Figure 174 is when the copy of carrying out the assignment bit map of being done when last check point is operated 74.
A check point data structure 84, the quantity of the check point operation that its record has been performed.When disk support 79 has been finished a mobile forward direction storage action and prepare another action of beginning, automatically perform the check point operation.At first, check point quantity 84 is set to 0.
Two pointer data structures, one first available block frame pointer 86 and a next available block frame pointer 88.When a check point operation took place, the first available block frame pointer 86 pointed to first available block frame.Next available block frame pointer 88 points to an available piece frame that will be used by next piece storage operation.Pointer 88 is updated on each piece storage operation.
When carrying out a disk and recover, recover the address of the first available block frame pointer 86 piece frame piece frame afterwards pointed by the piece storing process of server 26 " playbacks " after last check point.At first, two pointers all are configured to first not address of reserved block frame.
A transaction status log data structures 90, it contains the record tabulation relevant for transaction status.Each affairs is assigned with a universal unique identifier, is called Tid here.The record of daily record 90 comprises Tid and a tabulation 94 of participating in the TSLD of these affairs is associated with Tid.Log record comprises three parts: record type, transaction identifiers and tabulate 94.Describe record type, beginning affairs, the submission affairs that enter daily record 90 below in detail and stop affairs.
Above-described each data structure parts, be that conversion table 72, auxiliary conversion table 172, assignment bit map 74, work allocation bitmap 174, check point quantity 84, the first available block frame pointer 86, next available block frame pointer 88 and transaction status daily record 90 are safeguarded as TSLD interface 35 in primary memory 70, and in a check point operating process, be stored for TSLD28.
When a new piece frame is needed to store data, distributed by free block frame available from disk 62 according to 74, one piece frames of assignment bit map, and work allocation bitmap 184 is updated to reflect new distribution.In order to select which piece frame to be written into, server 26 writes down the value of next available block frame pointer 88.Server 26 uses this value with assignment bit map 74, is positioned with the next available block frame of the address before the last piece frame that once is written into.Like this, disk support 79 moves with mobile forward direction storage action, is called up to a check point operation.The temporary patent application 60/176507 of reference and be transferred to the same assignee of the present invention has been described the further details of relevant mobile forward direction storage means in the patented claim " Enhanced Stable Disk Storage (enhancing fixed disk storage) " that on July 6th, 2000 submitted in front.
When a client starts affairs on a plurality of TSLD, among the customer selecting TSLD one, preferably selecting has the TSLD of minimum sign as main TSLD.Main TSLD comes work as the register of transaction status.Describe as following, the record of transaction status daily record 90 preferably has three types.
The such fact of beginning transaction journal 92 record, promptly affairs Tid has begun and has been listed in the tabulation 94 of the TSLD that participates in these affairs.TSLD96 in the tabulation 94 is the main TSLD that selects for affairs Tid.Symbol for beginning transaction journal 92 used herein is:
Begin(Tid；TSLD 1，TSLD 2...TSLD m)
TSLD wherein 1Be main TSLD, TSLD 2TSLD mBe other TSLD that participates in affairs Tid.
Submit the such fact of transaction journal 102 records to for one, promptly affairs are submitted.The tabulation 94 that each submits to transaction journal 102 to comprise the sign Tid of affairs and participate in the TSLD of these affairs.Symbol for submission transaction journal 102 used herein is:
Commit(Tid；TSLD 1，TSLD 2...TSLD m)
One stops the such fact of transaction journal 112 records, and promptly affairs are terminated.Each stops the tabulation 94 that transaction journal 112 comprises the sign Tid of affairs and participates in the TSLD of these affairs.Symbol for termination transaction journal 112 used herein is:
Abort(Tid；TSLD 1，TSLD 2...TSLD m)
As described below, preferably each begins 92 of transaction journals and is stored in the transaction journal for the main TSLD of affairs Tid, and is submitted or stop up to affairs.
Preferably each submits to transaction journal 102 to be stored in the transaction status daily record 90 of main TSLD of affairs Tid.Preferably also do not know the interests of the TSLD that affairs are submitted, after in specific T SLD, having been upgraded, submit to transaction journal to be retained in the transaction status daily record of specific T SLD about the data structure of affairs Tid for other.In a specific main TSLD, submit to transaction journal to replace the beginning transaction journal.
Preferably each stops in the transaction journal 90 of main TSLD that transaction journal 112 is stored in affairs Tid.Preferably also do not know the interests of the TSLD that affairs are terminated for other, after having been upgraded in specific T SLD about the data structure of affairs Tid, each stops transaction journal and is retained in the transaction status daily record of specific T SLD.In a specific main TSLD, stop transaction journal and replace the beginning transaction journal.
In non-main TSLD, preferably related termination affairs and the submission transaction journal of this non-main TSLD remained into next check point always.
Fig. 3 is according to a preferred embodiment of the invention, is stored in block structured synoptic diagram on the disk by TSLD28.Data are used as piece 200 and store in the specific frame 82.Piece 200 comprises a client portion 202, uses its content-data of storage by client 202, and an afterbody 204 comprises the territory of being used by TSLD28.Afterbody 204 comprises following territory:
Check point number field 206 on disk has wherein been stored the currency of check point quantity 84.This is the check point quantity when the piece storage operation.
The LBA of piece 200 has wherein been stored in a LBA (Logical Block Addressing) (LBA) territory 208.
The mark of having described piece 200 types has wherein been stored in a block type territory 210.The type of piece 200 is described below.
The sign Tid of the affairs that piece 200 participating in has wherein been stored in Transaction Identifier territory 212.Alternately, if when piece is stored, piece 200 then is set to 0 with territory 212 not participating in affairs.
The sign TSLD of the main TSLD of the affairs that piece 200 participating in has wherein been stored in main TSLD territory 214 pAlternately, if when piece is stored, piece 200 then is set to 0 with territory 214 not participating in affairs.
The piece of following type can be supposed in block type territory 210:
A data block type, wherein the content-data from client 22 is stored in the customer data territory 202, and wherein stored frame of piece located by transaction table 72, and wherein piece is not the part of open affairs.
The soft block type of writing, wherein the content-data from client 22 is stored in the customer data territory 202, and the not submission piece frame that data are stored in is pointed to by table 72, but the submitted old piece frame that content was stored in is not released being used by client 22, but is pointed to by table 172.Two piece frames all are labeled distribution in assignment bit map 74.
A begin block type wherein identifies particular transaction Tid and stores in the territory 202 with the sign of the TSLD that participates in affairs Tid.This such fact of record, promptly affairs Tid begins, and begin block takies its piece frame separately, up to next check point operation.After the check point operation of a success, as mentioned above, discharge occupied frame.
Submit block type to for one, wherein particular transaction is identified Tid and store in the territory 202 with the sign of the TSLD that participates in affairs Tid.A specific submission piece writes down such fact, and promptly affairs Tid is submitted, and takies its piece frame separately, up to next check point operation.After the check point operation of a success, as mentioned above, discharge occupied frame.
One stops block type, wherein particular transaction is identified Tid and stores in the territory 202 with the sign of the TSLD that participates in affairs Tid.A specific termination piece writes down such fact, and promptly affairs Tid is terminated, and takies its piece frame separately, up to next check point operation.After the check point operation of a success, as mentioned above, discharge occupied frame.
Fig. 4 has provided according to a preferred embodiment of the invention, is included in the process flow diagram of the step in the soft write operation 220.Soft write operation 220 has the main TSLD of primTSLD value to store in each the piece frame that is upgraded by affairs Tid the content of affairs Tid.Symbol for soft write operation 220 used herein is:
Soft-Write(Tid，primTSLD，i，contents)
Hypothesis TSLD28 participates in one of TSLD below, and supposes that operation 220 is started by client 22.Operation 220 is also relevant with a LBAi with content, and the old content of not losing LBAi.Soft write operation 220 is as an interim write operation action, and it can be submitted or stop afterwards, thereby moves as first section of two-phase commitment protocol, and this was described in background of invention.
In checking step 222, check conversion table 72 whether searching corresponding to LBAi, TT (i) is provided with its mark.If mark has been set up, then will return to client 22 below with reference to the open transaction error code that Figure 16 will describe its function.If mark is not set up, then in scanning step 224, assignment bit map 74 is scanned piece frame first available block frame afterwards of following record in next available block frame 88 to search.If in availability inspection step 226, do not find available piece frame, then carry out a check point operation as described below.If a piece frame is found, suppose that then its physical address is " a. ".
In storing step 228, check point quantity, LBAi, Transaction Identifier Tid and main these values of TSLD primTSLD are written in the afterbody territory separately 206,208,212 and 214 (Fig. 3) of the piece in the piece frame that writes down in the next available block frame 88.Will be corresponding to block type, the promptly soft value of writing block type is written in the afterbody territory 210 of piece frame.In addition, in the data field 202 of the data content write-in block frame of the Tid that client 22 is provided.
In step of updating 230, in storer 26, carry out following distribution:
In work allocation bitmap 174, A (next available block frame) ← 1;
In auxiliary conversion table 172, (ATT (i), Trans ID) ←＜TT (i), Tid 〉;
In conversion table 72, TT (i) ← next available block frame, and the mark of TT (i) is provided with in tag field 73; And
Next available block frame 88 values of being configured to " a. " of data structure
Should be appreciated that soft write operation 220 writes data and afterbody numeric field data in the piece of next available block frame 88 as the part of the affairs that have been done.But, the specific frame that is used for holding the old content of i is not changed, and the allocation bit by this specific frame is set to 0 and do not discharge this piece frame, and so if necessary, old content still can be accessed.
Fig. 5 has provided according to a preferred embodiment of the invention, is included in the process flow diagram of the step in the soft read operation 240.As a part of not finishing affairs, soft read operation 240 reads in stored data content in the soft write operation 220.Symbol for soft read operation 240 used herein is:
Soft-Read(i)
In checking step 241, check conversion table 72 to search corresponding to LBAi, whether the item of TT (i) is provided with its mark.If mark is not set up, then an open transaction error code is returned to client 22.If mark is set up, then in read step 242, such as a requester requests of client 22 in conversion table 72, the data content 202 of the piece in the piece frame that points to by TT (i).Data content 202 is returned to the requestor, and soft then read operation 240 stops.
Fig. 6 has provided according to a preferred embodiment of the invention, is included in the process flow diagram of the step in the read operation 250.The data content that operation 250 is used to read the logical block i in the piece frame of having finished affairs, and this content returned to a requestor such as client 22.Symbol for read operation used herein is:
Read(i)
In first step 254, the data content 202 of the piece in the piece frame that is pointed to by TT (i) in conversion table 72 is read from TSLD.By check TT (i) mark whether be set up, operation 250 continues.If mark is set up, show that the affairs that comprise TT (i) also are not done, then an open transaction error signal is returned to the requestor who has before read the piece frame.If the mark of TT (i) is not set up, show that piece is not included in the open affairs, then by the piece frame is returned to the user, operation 250 is finished.Operating 250 then stops.
Fig. 7 has provided according to a preferred embodiment of the invention, is included in a process flow diagram in the step that begins to operate in 260.Operation 260 storages are called a piece of begin block here, and it has write down the details of affairs on a main TSLD.As top described with reference to figure 2, operation 260 is storage beginning transaction journal 92 also, Begin (Tid; TSLD 1, TSLD 2... TSLD m).Suppose that these affairs have a sign Tid, and comprise following TSLD:TSLD 1, TSLD 2... TSLD mTSLD preferably 1Main TSLD (primTSLD) work as affairs Tid.A transaction coordination person, hypothesis is client 22 here, starts main TSLD, i.e. TSLD 1On begin to operate 260.Used hereinly be for the symbol that begins to operate 260:
Begin(Tid；primTSLD，TSLD 2，TSLD 3...TSLD m)
In recording step 262, the TSLD of the participation of sign Tid and affairs, i.e. TSLD 1, TSLD 2, TSLD 3... TSLD mBe recorded in the transaction journal 90.In scanning step 264, assignment bit map 74 is scanned piece frame first available block frame afterwards of following record in next available block frame 88 to search.If check in the step 266 in availability, do not find the available block frame, then carry out the check point operation.If piece frame subsequently is found, suppose that then its physical address is " a. ".
In storing step 268, with check point quantity, Transaction Identifier Tid and TSLD 1(primTSLD) these values write in the afterbody territory separately 206,212 and 214 (Fig. 3) of the piece that is recorded in the piece frame in the next available block frame 88, suppose that here the physical address of this piece frame is " p. ".Will be corresponding to block type, promptly in the afterbody territory 210 of the value write-in block of begin block type.In logical address territory 208, do not do any.Equally, in storing step 268, the participation TSLD of affairs, i.e. TSLD 1, TSLD 2... TSLD mQuilt is record in the data field 202 of the piece of piece frame " p. ".
In step of updating 270, in storer 26, next available block frame 88 values of being set to " a. " of data structure.Operating 260 then stops.
Should be pointed out that work allocation bitmap 174 is not updated in operation 260, promptly A (p) keeps being set to 0.Like this, piece frame " p " keeps being labeled as the free time.But, when operation 260 is finished, " a " greater than " p ", like this, the piece of physical address greater than " p. " will be always sought in the operation after operation 260.Like this, in the current scanning process of disk support 79, piece frame " p " will not be stored into.When carrying out a check point operation, current scanning stops, and in this check point operating process, all in the piece frame " p " have related parameter all to be stored in the disk 62.After the check point operation, piece frame " p " becomes available for further storage operation.
Fig. 8 has provided according to a preferred embodiment of the invention, is included in the process flow diagram of a step in the submit operation 280.Operation 280 should all successfully just be performed after the termination at all soft write operations 220 relevant with given affairs, and the feasible interim write operation of being done becomes permanent.Carry out on each TSLD that should in affairs Tid, be comprised for the operation 280 of soft write operation 220.Symbol for submit operation 280 used herein is:
Commit(Tid；primTSLD，TSLD 2，TSLD 3，...TSLD m)
As described below, step 284, one of 286 and 288 storage submit to piece to disk 62.The submission piece of being stored comes work as the temporary storage of a submit operation, till having carried out a check point operation subsequently.
In scanning step 284, assignment bit map 74 is scanned piece frame first available block frame afterwards of following record in next available block frame 88 to search.If in availability inspection step 286, do not find the available block frame, then carry out a check point operation.If found a piece frame subsequently, suppose that then its physical address is " a. ".
In storing step 288, check point quantity, Transaction Identifier Tid and TSLD 1(primTSLD) these values are written in the territory of afterbody separately 206,212 and 214 (Fig. 3) of the piece in the piece frame that writes down in the next available block frame 88, suppose that here the physical address of this piece frame is " p. ".Corresponding to block type, promptly the value of a submission block type is written into the afterbody territory 210 of piece frame.In logical address territory 208, do not do any.Equally, in storing step 288, the participation TSLD of affairs, i.e. TSLD 1, TSLD 2... TSLD mQuilt is record in the data field 202 of the piece of piece frame " p. ".
In step of updating 290, in storer 26, next available block frame 88 values of being set to " a. " of data structure.
In scanning step 292, for the item that comprises Transaction Identifier Tid, auxiliary conversion table 172 is scanned.For each found item,
Corresponding bits in the work allocation bitmap 174 is set to 0, and the piece frame that is pointed to by item becomes available like this.
In conversion table 72, be eliminated corresponding to those each the mark that in auxiliary conversion table 172, finds.
Item in the auxiliary conversion table 172 is removed, so that to simplify form maintenance table 172.
In recording step 294, if TSLD28 is main TSLD, i.e. TSLD 1,, submit to record institute to be replaced by of the TSLD of affairs Tid and participation, submit operation 280 terminations then then as the beginning transaction journal of transaction journal 90 performed in the recording step 262 (Fig. 7).If TSLD28 is not main TSLD, then after step 292, operation 280 stops.
Should be appreciated that piece frame " p " is only occupied next check point operation, this piece frame becomes available afterwards.
Fig. 9 has provided according to a preferred embodiment of the invention, is included in the process flow diagram of the step in the terminating operation 300.If one in the soft write operation 220 relevant with given affairs has been failed, then should executable operations 300.Operation 300 is withdrawed from from operate the interim write operation of carrying out 220, and before executable operations 220, makes the content of piece frame become available.As for soft write operation 220, operation 300 is carried out by on each TSLD in being included in affairs Tid.Symbol for terminating operation 300 used herein is:
Abort(Tid；primTSLD，TSLD 2，TSLD 3，...TSLD m)
Step 304,306,308 and 310 stops piece with one and stores disk 62 into.The termination piece of being stored comes work as the temporary storage of terminating operation, till having carried out a check point operation subsequently.In operation, step 304,306 and 310 is substantially similar to aforesaid step 284,286 and 290.In storing step 308, check point quantity, Transaction Identifier Tid and TSLD 1(primTSLD) these values are written in afterbody territory separately 206,212 and 214 (Fig. 3) of the piece in the piece frame that writes down in the next available block frame 88, suppose that here this piece frame has a physical address " q. ".Corresponding to block type, the value that promptly stops piece is written into the afterbody territory 210 of piece.In logical address territory 208, do not do any.Equally, in storing step 308, the participation TSLD of affairs, i.e. TSLD 1, TSLD 2... TSLD mBe recorded in the data field 202 of the piece in the piece frame " q. ".
In scanning step 312, for the item that comprises Transaction Identifier Tid, the auxiliary conversion table 172 of scanning.For each item that finds, conversion table 72 is used to locate corresponding piece frame.Corresponding bits in work allocation bitmap 174 is reset then.In step 312, after bit was reset, conversion table 72 was updated equally, so that found item points to the item that finds now in auxiliary conversion table 172.At last, with each removing in the auxiliary conversion table 172, and the mark of each corresponding entry in the conversion table 72 removed.
In recording step 314, if TSLD28 is main TSLD (TSLD 1), then as the beginning transaction journal of the performed transaction journal 90 of recording step 262 (Fig. 7) by an affairs Tid and participate in TSLD the termination record replace, terminating operation 300 stops then.If TSLD28 is not main TSLD, then after step 312, operation 300 stops.
Should be appreciated that piece frame " q " is only occupied uses till the operation of next check point, and this piece frame becomes available afterwards.
Figure 10 A has provided according to a preferred embodiment of the invention, is included in the process flow diagram of a step in the query manipulation.For each TSLD in system 20 or the system 40, query manipulation 320 is checked transaction status daily records 90 states about a particular transaction Tid, and the information about these affairs known to inciting somebody to action returns to a requestor such as client 22.Symbol for query manipulation 320 used herein is:
Query_State(Tid)
In first inquiry 322, audit log 90 is to see whether the opening entry for particular transaction exists.If opening entry exists, show that the TSLD that is inquired about is that main TSLD and these affairs Tid are also underway, promptly be open, then return one and have the open state that returns of participating in TSLD tabulation (from opening entry, drawing).Show that these affairs are TSLD open and that listed other participation.
If an opening entry does not exist, then operate 320 and proceed to second inquiry 324, check that wherein auxiliary conversion table 172 is to see whether this table comprises the item for affairs Tid.If this table comprises the one or more items for Tid, show that these affairs are underway, then return an open to the outside world statement.
If the answer is in the negative for inquiry 324, then operate 320 and proceed to the 3rd inquiry 326, check that wherein transaction journal 90 is to see whether the termination record for particular transaction exists.If one stops record and exists, then return one " termination " statement.
If the answer is in the negative for inquiry 326, then operate 320 and proceed to the 4th inquiry 328, check that wherein transaction journal 90 is to see whether the submission record for affairs exists.If one is submitted to record to exist, then return one " submission " statement.If one is submitted to record not exist, then return one " the unknown " statement, and exist 320 to stop.
Figure 10 B has provided according to a preferred embodiment of the invention, from the constitutional diagram 330 of the possible outcome of the operation on the specific T SLD 320.Shown in Figure 33 0, particular transaction Tid can be terminated in state 334, a submitted state 336 or a unknown state 338 an open state 332, one and exist.
If the main TSLD for an affairs Tid responds a query State with " the unknown ", and another non-main TSLD responds this inquiry with the open to the outside world state, can infer that then these affairs are submitted.This situation occurs in when these affairs successful execution in main TSLD, but in non-main TSLD successful execution submit operation not.
Figure 11 has provided according to a preferred embodiment of the invention, is included in the process flow diagram that there is the step in 340 in check point.Though below about operating 340 description application-specific in TSLD28 (Fig. 2), the identical step in the check point operation of carrying out separately that all TSLD in system 20 or the system 40 follow basically and describe below.Check point operation 340 writes disk 62 with the value of holding in the TSLD interface 35 in the volatile memory 70, and its intermediate value is by permanent storage.Preferably check point is operated the 340 piece frame 82e that write on the disk 62 that is pre-assigned to this operation.Preferably not with identical space among the check point data write-in block frame 82e, but write, thereby previous check point data are not write immediately or are wiped with a kind of interactive mode.Check point operation 340 will copy to from the data structure of TSLD35 in the disk 62 with all period interval, and in turkey, TSLD28 is quick-recovery soon like this.Can carry out check point operation 340 by TSLD28 at any time, but must be when when having the available block frame around the next available block frame 88, not carrying out.
In initial step 342, all data structures in the TSLD interface 35 are lockable, and TSLD28 stops to support the operation except that the check point operation like this.In second step 344, check point operation 340 values that read from first available block frame 86 of interface 35.Piece frame 86 be free time of from work allocation bitmap 174, determining and by the piece frame relevant with minimum disk sector address.In increasing step 346, the value of check point quantity 84 is increased, and the value that is increased is read.
In storing step 348a, 348b, 348c, 348d and 348e, operation 340 with conversion table 72, duplicate conversion table 172, work allocation bitmap 174, first available block frame 86, the check point quantity 84 that is increased and transaction status daily record 90 and be written among the piece frame 32e.
In the first reassignment step 350, the value of next available block frame 88 is become value as first available block frame 86 that finds in the step 344 by assignment, and its next available block frame is used to a mobile forward direction storage operation subsequently.In the second reassignment step 352, assignment bit map 74 is become the value of work allocation bitmap 174 by assignment.Operate data structure that is lockable and termination in the 340 release TSLD interfaces 35 then, so TSLD28 utilizes the data structure recovery operation that is updated.
When check point operation 340 ran on the specific T SLD, TSLD stopped other operation.In order to reduce influence for a system user, can be in minor time slice executable operations 340.In above-mentioned interim patent sector and the patented claim of above-mentioned being entitled as " Enhanced Stable DiskStorage (enhancing fixed disk storage) ", the check point operation that is substantially similar to operation 340 has been described.In addition, explained in detail that also being used for carrying out check point in minor time slice operates to reduce the method to the system user influence.Those of skill in the art can adopt method in the patented claim that is entitled as " Enhanced Stable Disk Storage (enhancing fixed disk storage) " with executable operations in minor time slice 340.
Figure 12 has provided according to a preferred embodiment of the invention, is included in the process flow diagram for the step in the coordinated manipulation of affairs Tid.Here suppose to start, and the TSLD of these affairs of hypothesis participation is TSLD by client 22 as the transaction coordination person 1, TSLD 2... TSLD m
In initial step 362, a conduct among the TSLD that client 22 selects to participate in is for the main TSLD work of affairs Tid.Know the TSLD of symbol or least activity or at random carry out this selection by selecting to have lowest bid.Here suppose TSLD 1Be main TSLD (primTSLD).
In beginning affairs step 364, client 22 will begin to operate 260 (Fig. 7), Begin (Tid; PrimTSLD, TSLD 2, TSLD 3... TSLD m) submit to TSLD 1, and in first waiting step 365, for one from TSLD 1Wait for a predetermined amount of time about beginning to operate the affirmation that completed successfully.If do not confirm, then client 22 submits terminating operation 300, Abort (Tid to; PrimTSLD, TSLD 2, TSLD 3... TSLD m) to TSLD 1And stop this affairs.
Begin to operate 260 affirmations that completed successfully if client 22 receives, then write in the step 366 soft, client 22 is with one or more soft write operations 220 (Fig. 4), Soft-Write (Tid; PrimTSLD, i contents) submits among the TSLD of participation each.Client 22 will descend column data to offer the TSLD of each participation: Transaction Identifier Tid, main TSLD identify (TSLD 1), interrelated logic block address and want stored relevant data block content.
In second waiting step 368, client 22 waits for that a predetermined amount of time is to receive from each the positive acknowledgment that is completed successfully about each soft write operation among the TSLD that participates in.This continues for each soft write operation.
If in the section, all do not receive positive acknowledgment from the TSLD of all participation at the fixed time, then in stopping step 370, client 22 is terminating operation 300, Abort (Tid; PrimTSLD, TSLD 2, TSLD 3... TSLD m) submit to each among the TSLD of participation.Do not completed successfully if main TSLD has to confirm to operate 300, then wait for finish the submission of affirmation for each after, client 22 and then again with an absorption (sticky) terminating operation (below will explain), StickyAbort (Tid; PrimTSLD, TSLD 2, TSLD 3... TSLD m) submit among the non-main TSLD of participation each in a single day receive an affirmation, if or the TSLD of all participation is inquired about and neither one is made response, then stop coordinated manipulation 360.
If in the section, received positive acknowledgment from the TSLD of all participation at the fixed time, then in submitting step 372 to, client 22 is submit operation 280, Commit (Tid; PrimTSLD, TSLD 2, TSLD 3... TSLD m) submit to the main TSLD of affairs.If main TSLD has confirmed this submit operation, operate each among the TSLD that is sent to participation then.Do not completed successfully if TSLD has to confirm to operate 280, then wait for finish the submission of affirmation for each after, client 22 and then again with an absorption submit operation (below will explain), Stickycommit (Tid; PrimTSLD, TSLD 2, TSLD 3... TSLD m) submit to each among the non-main TSLD of participation.In case receive an affirmation, if or the TSLD of all participation is inquired about and neither one is made response, then stop coordinated manipulation 360.
It is identical with submit operation basically to adsorb submit operation, and corresponding absorption submits to record identical with the submission record basically.Absorption submit operation and record are applied to non-main TSLD.As the result of absorption submit operation, an absorption submits to record to be added in the transaction status daily record of non-main TSLD.Can not submit to record to remove an absorption, unless indicate clearly by a client from the transaction status daily record of a non-main TSLD.Top description also is applied to adsorb the relation between terminating operation and the terminating operation, and adsorbs the relation between stopping writing down and stopping writing down.
Should be appreciated that coordinated manipulation 360 attempts to submit to or stop affairs Tid, and the result of recording operation in the participation TSLD of affairs at least one.
Figure 13 has provided according to a preferred embodiment of the invention, is included in the process flow diagram of the step in first recovery operation 373.After from the storage server fault such, recovering such as the server system 20 or the system 40 26, executable operations 373.As described in detail later, operation 373 requires scanning assignment bit map 74.(patented claim of above-mentioned being entitled as " Enhanced StableDisk Storage (enhancing fixed disk storage) " has provided the detailed description of the alternative method of recovering from the fault of an on-fixed magnetic disk memory, those of skill in the art can use this description recovers the server such as server 26 from the system failure.The time of this alternative method cost is the linear function of the quantity of the piece write since the check point since last time.)
In first read step 374, the value that has been stored in the assignment bit map 74 on the disk 62 in check point operation last time 340 is read back in TSLD interface 35 in assignment bit map 74 and the work allocation bitmap 174.
In second read step 375, the value that is stored in first available block frame 86 on the disk 62 in check point operation last time 340 is read back in first available block frame 86 and the next available block frame 88.
Get in the step 376 in third reading, the value in check point operation 340 last time in stored conversion table 72, auxiliary conversion table 172, check point quantity 84 and the transaction status daily record 90 is read back in TSLD interface 35 in the data structure separately.
In the 4th read step 377, the piece that is stored in the piece frame address that is provided by next available block frame 88 is read.
In step of updating 378, the check point quantity 84 in checking TSLD interface 35 corresponding to the piece that is read for each be stored check point quantity the time,
1. (from step 377 beginning) upgrades all related data structures the TSLD interface 35 according to the piece that was read last time.The TSLD operation of promptly resetting and when the storage block frame, occurring.
2. by first bit of an available block frame of in assignment bit map 74, having searched mark forward, and in this piece frame, read piece, locate the next piece frame that will be stored in.
3. continue step 378a and 378b, ignore any that to read owing to fault sector 64, up to till forward direction is finished scanning assignment bit map 74.
Operating 373 then stops.
Be to be understood that, the TSLD that utilizes " playback " to have carried out since the check point operation since last time operates this process, first recovery operation 373 turns back to the data structure on the TSLD interface 35 value before their faults basically fully, except by the caused localized mistake of the piece that can not be read.Realize this playback procedure by the one or more parameters that write disk 62 from check point operation last time, so that set up the piece frame position that behind aforesaid operations, writes.
Figure 14 has provided according to a preferred embodiment of the invention, is included in the process flow diagram of the step in second recovery operation 380.Operation 380 is preferably carrying out after one hypothesis is the operation 373 that finishes on the server that is resumed of server 26 here, so that determine the state of open affairs.
In the first step 382 of recovery operation 380, server 26 scans its transaction status order will 90 for the beginning transaction journal, and this record has indicated wherein that TSLD28 is the open affairs of main TSLD.For each such affairs, in query steps 384, TSLD28 submits or stops record to for one, checks the TSLD of each participation of affairs.In step of updating 386, TSLD 28 upgrades its transaction journal 90 according to the result of step 384.First recovery operation 380 stops then.Should be appreciated that by executable operations 380, TSLD in a state, in this state, its can be enough the lastest imformation response from the inquiry that is included in other TSLD in the affairs that TSLD28 wherein is main TSLD.
Figure 15 has provided according to a preferred embodiment of the invention, is included in the process flow diagram of the step in the 3rd recovery operation 390.As for recovery operation 380, preferably by server 26 after from fault, recovering, and after having carried out operation 380, carry out recovery operation 390.In first step 392, server 26 its rebuilt as mentioned above auxiliary conversion tables 172 of scanning it seems it is open affairs to search.For each such affairs, the step below server 26 is carried out.In query steps 394, the soft piece of writing that server 26 is located corresponding to particular transaction, and inquire about its transaction journal 90 to find the state of affairs.If transaction journal 90 provides the state of affairs, server 26 data structure in the updated stored device 70 and continue next open affairs on request then.
If transaction journal 90 can not provide the state of affairs, then in second query steps 396, inquired about as the main TSLD of the affairs of afterbody territory 214 regulations of the piece frame of affairs.Server 26 will be included in for the Query Result of main TSLD in the data structure of TSLD interface 35.When all open affairs that find in the step 392 all by step 394 and/or step 396 processed after, operation 390 stops.
Figure 16 has provided according to a preferred embodiment of the invention, is included in the affairs complete operation of being carried out by the client 400.The result of client's fault when typically, this is as the execution affairs.Operation 400 preferably receives the execution when returning of an open transaction error by the client in system 20 or the system 40 when it attempts to read specific frame in the particular transaction.(as mentioned above, for example in the inspection step 222 of soft write operation 220.), suppose that the client is client 22 here, suppose that the piece frame is the piece frame among the TSLD28.Open transaction error indicator dog frame 64a is participating in open affairs.Operation 400 makes the participator of open affairs can set up the consensus of whether submitting to or stopping these affairs, thereby finishes this affairs.
In first step 402, client 22 carries out soft read operation 240 (Fig. 5) on piece frame 64a, i.e. Soft-Read (i), and the sign of reception Transaction Identifier Tid and main TSLD.In first query steps 404, client 22 carries out query manipulation 320 (Figure 10 A) on main TSLD, i.e. Query_State (Tid), and come work according to the response that receives.
If not response, then because main TSLD is not moving, client 22 can not finish these affairs.In this case, in waiting step 406, client 22 waits for that main TSLD is sent to an operating server.
If in first response of step 408, main TSLD responds with one " the unknown " statement, and then client 22 carries out submit operation 280, i.e. Commit (Tid on specific T SLD; PrimTSLD, TSLD 2, TSLD 3... TSLD m), going up the client from this TSLD and receive open transaction error signal, this specific T SLD is TSLD28 in this case.
If in second response of step 410, main TSLD comes together to respond with the tabulation of other TSLD that participates in these affairs with an open to the outside world statement, then in second query steps 412, client 22 and then carry out query manipulation 320 on each in the TSLD that participates in again.
If any one from the TSLD that participates in receives a clear and definite response, i.e. " submission " or " termination " answer, then in the 3rd response of step 414, client 22 carries out submit operation 280 or terminating operation 300 separately, i.e. Abort (Tid on the TSLD of remaining participation; PrimTSLD, TSLD 2, TSLD 3... TSLD m).
If in second response of step 412, the TSLD of all participation returns an open to the outside world and answers, then in the 4th response of step 416, just in case after operation 400 beginnings a variation is being arranged, client 22 manages to read initial piece frame, i.e. piece frame 64a once more.If do not change, then client 22 executive termination operation 300 on the TSLD that participates in.
If in second response of step 412, a not response among the TSLD of participation, then in the 5th response of step 418, the response that client 22 waits for from specific T SLD is because the not TSLD of response that may be specific has desired answer.After step 406,408,414,416 or 418, operation 400 stops.
Should be appreciated that in the addressable institute of TSLD28 is free specific frame 82 of TSLS28 or do not participate in affairs is perhaps participated in, so that transaction table 72 can should be participated in by marks.Identical situation is applied to other addressable TSLD and their conversion tables separately in system 20 or the system 40.Like this, in affairs complete operation 400, client 22 can read the piece content of a piece frame on the TSLD and the piece content of the piece frame on another TSLD, and system 20 or system 40 seem inconsistent like this.But, even in this case, the client also knows the state of system and can draw the state of not finishing affairs about any.
In order to minimize the size of transaction status daily record 90, when no longer needing to write down, record need be removed from daily record.As mentioned above, in main TSLD, each begins transaction journal and is submitted to record or a termination record to replace by one.After the preset time interval, the submission record on a main TSLD can be eliminated, because no longer needed them.(as mentioned above, if known affairs are open on a non-main TSLD, and main TSLD responds with " the unknown " when by when inquiry, then suppose affairs submit to).In case main TSLD has confirmed the TSLD of all participation of relevant issues and has known the state of affairs that then the record of the termination on main TSLD can be eliminated.
In case inquired about a non-main TSLD about these affairs for the main TSLD of affairs, then this non-main TSLD just can remove the submission of a particular transaction or stop record, because when inquiring about, main TSLD has write down the state of affairs.Therefore the most preferably, coordinate as described above at an intrasystem TSLD, so that unwanted transaction journal is removed from transaction status daily record 90.
Used a plurality of servers of a plurality of TSLD and these TSLD of management although should be appreciated that system 20 and system 40, principle of the present invention also is applicable to the server of a TSLD of management.In addition, should be appreciated that when a plurality of concurrent transactions are performed that principle of the present invention is applicable to the recovery of described a plurality of concurrent data transactions and finishes on one or more TSLD, because these a plurality of affairs are mutual separate records and recovery basically.
It should also be understood that, scope of the present invention comprises except above-described and being used for after a fault of server, method from consecutive data block outside the method for restore data wherein indicates one or more parameters of the position of a consecutive data block to be written into a memory devices.The example of these methods is described with being entitled as in " Enhanced Stable DiskStorage (enhancing fixed disk storage) " at temporary patent application 60/176507.
It is also understood that the preferred embodiments of the invention to be used as computer software product and provide.A kind of like this product can be used as an electronic data transfer or to provide such as the tangible form on the such computer-readable medium of CD-ROM or the mixing of two kinds of forms.
Therefore, should be appreciated that above-mentioned preferred embodiment utilizes example to quote, and the invention is not restricted to be illustrated especially and content described above.But scope of the present invention comprises above-described various combination of features or sub-portfolio, and after the undocumented description, those skilled in the art can make variations and modifications of the present invention in having read previously described priori technology.

Claims (60)

1. be used for the equipment of support data affairs, it is characterized in that, described equipment comprises:
At least one memory devices is used for receiving as the data of consecutive data block with separately physical location storage therein; And
Control circuit, be configured when writing at least one memory devices, with at least some data blocks and indicated in consecutive data block the parameter of the physical location of data block afterwards to be sent at least one memory devices, and utilize at least one parameter to come the state of translation data consecutive data block to finish data transactions.
2. equipment as claimed in claim 1, wherein control circuit comprises at least one communication link that control circuit is connected at least one memory devices, wherein control circuit is by this at least one memory devices of at least one link management.
3. equipment as claimed in claim 1 comprises with control circuit linking to each other, and is used for the data transmission network of data transmission to control circuit.
4. equipment as claimed in claim 3, wherein data transmission network is connected at least one memory devices, and data are sent to this at least one memory devices, and wherein control circuit is managed at least one memory devices by this data transmission network.
5. equipment as claimed in claim 1 comprises a client who data is sent to control circuit.
6. equipment as claimed in claim 1, wherein control circuit comprises at least one volatile memory, be used for storing one or more data structures of the data that contain one or more characteristics of having indicated at least some data blocks, described data write at least one memory devices by control circuit, so that the content of this at least one volatile memory can be regenerated from least some data in the one or more data structures being stored in this at least one memory devices.
7. equipment as claimed in claim 6, wherein in the data structure comprises a conversion table, be used for LBA (Logical Block Addressing) with consecutive data block to be mapped to separately physical address, and sign which data block wherein is current is participating in an open data transactions.
8. equipment as claimed in claim 7, wherein in the data structure comprises an auxiliary conversion table, be used for for each open data transactions separately, will the LBA (Logical Block Addressing) relevant be mapped to the physical address separately of the consecutive data block that data transactions exists before beginning, and be mapped to a sign of this opening data transactions with open affairs.
9. equipment as claimed in claim 7, wherein control circuit writes consecutive data block with LBA (Logical Block Addressing) separately.
10. equipment as claimed in claim 7, wherein control circuit writes consecutive data block with the sign of data transactions.
11. equipment as claimed in claim 7, wherein control circuit writes consecutive data block with the sign of the register of the state of data transactions.
12. equipment as claimed in claim 6, wherein in the data structure comprises the assignment bit map of the availability of a mapping each in the consecutive data block.
13. equipment as claimed in claim 6, wherein a pointer value that comprises the position of a next available block in the consecutive data block in the data structure.
14. equipment as claimed in claim 6, wherein in the data structure comprises the pointer value of the position of first available block in the consecutive data block.
15. equipment as claimed in claim 6, wherein at least one memory devices includes the disk of a magnetic head, and wherein control circuit writes disk in the mode of a series of scannings on magnetic disk surface that make progress at folk prescription with data block.
16. as the equipment of claim 15, each in wherein a series of scannings has a check point quantity, and wherein in the data structure comprises the value of having indicated current check point quantity.
17. as the equipment of claim 15, wherein in the data structure comprises a work allocation bitmap, is used for when one or more scannings of magnetic head are finished the availability of each in the storage consecutive data block.
18. as the equipment of claim 15, wherein control circuit writes at least one memory devices with at least some data in one or more data structures when one or more scannings of magnetic head are finished.
19. equipment as claimed in claim 6, wherein in the data structure comprises a transaction status daily record that is used for storing at least one record in the data transactions, wherein at least one record comprises the sign of data transactions and the sign of participating at least one memory devices of this data transactions, and wherein at least one record comprises at least one in the group record of being made up of following content: the beginning affairs, submit affairs to and stop affairs, they write down respectively affairs begin, submitted and stopped.
20. equipment as claimed in claim 1, wherein control circuit writes consecutive data block with an afterbody, and this afterbody comprises one or more data fields of the use of having indicated each data block.
21. as the equipment of claim 20, wherein one or more data fields comprise the sign of data transactions and the sign of data transactions register, wherein the register of data transactions comprises at least one memory devices one.
22. as the equipment of claim 20, wherein one or more data fields comprise a logical address of data block.
23. as the equipment of claim 20, wherein one or more data fields comprise the sign of having indicated the data transactions state.
24. equipment as claimed in claim 1 comprises the Magnetic Disk Controller of controlling at least one memory devices, wherein Magnetic Disk Controller can recover the content of a volatile memory from the data that write at least one memory devices by control circuit.
25. equipment as claimed in claim 1, wherein data transactions comprises a plurality of concurrent data transactions.
26. equipment as claimed in claim 1, wherein at least one memory devices comprises the disk by a server admin.
27. equipment as claimed in claim 1, wherein at least one memory devices comprises a plurality of memory devices.
28. equipment as claimed in claim 1, wherein at least one parameter comprises the transaction status daily record of one or more states of record data affairs.
29. equipment as claimed in claim 1, wherein at least one parameter comprises at least a portion of distributed transaction state daily record of one or more states of record data affairs.
30. equipment as claimed in claim 1, wherein at least one memory devices be changed to mobile so that by the operation of other control circuit to recover to be stored in the information at least one memory devices.
31. a method that is used for carrying out data affairs comprises:
In at least one memory devices, receive as the data of consecutive data block, at least some data blocks each and indicated in consecutive data block at least one parameter of the physical location of data block afterwards; And
In response at least one parameter, change the state of the data at least one memory devices, to finish this data transactions.
32., wherein at least one memory devices, receive data and comprise by a data transmission network data are sent at least one memory devices as the method for claim 31.
33. as the method for claim 32, wherein transfer data packets is drawn together at least one memory devices is connected to data transmission network.
34. as the method for claim 32, wherein transfer data packets is drawn together:
By control circuit at least one memory devices is connected to data transmission network; And
Control circuit is connected to data transmission network.
35., wherein receive data and comprise from the client and receive data as the method for claim 31.
36. method as claim 31, be included in one or more data structures of storing the data that contain one or more characteristics of having indicated at least some data blocks at least one volatile memory, and at least some data in the data structure are write at least one memory devices, so that regenerate at least some data in can the one or more data structures from be stored at least one memory devices of the content at least one volatile memory.
37. method as claim 36, wherein store one or more data structures and comprise conversion table of storage, being used for LBA (Logical Block Addressing) with consecutive data block is mapped to separately physical address, and sign which data block wherein is current is participating in an open data transactions.
38. method as claim 37, wherein store one or more data structures and comprise auxiliary conversion table of storage, be used for for each open data transactions separately, the LBA (Logical Block Addressing) relevant with this opening data transactions is mapped to the physical address separately of the consecutive data block that existed before this opening data transactions begins, and is mapped to the sign of this opening data transactions.
39., wherein receive data and comprise separately logical address is write consecutive data block as the method for claim 37.
40., comprise and use the certain data block in conversion table location, with reading of data from this certain data block as the method for claim 37.
41., wherein store one or more data structures and comprise assignment bit map of having indicated the availability of each in the consecutive data block of storage as the method for claim 36.
42., wherein at least some data are write at least one memory devices and comprise and use the following step that data are write in the consecutive data block one as the method for claim 36:
Scan one or more data structures to determine the physical address of an available block at least one memory devices;
At least some contents of data and one or more data structures are write physical location; And, upgrade one or more data structures in response to the physical location that this is determined.
43., wherein scan one or more data structures and comprise whether LBA (Logical Block Addressing) of inspection is included in the open data transactions as the method for claim 42.
44. as the method for claim 42, at least some contents that wherein write in data and the one or more data structure comprise step:
Write value corresponding to a soft write operation; And
Write as one sign at least one memory devices of a transaction status register, to carry out an interim write operation.
45. method as claim 42, wherein writing at least some contents in data and the one or more data structure comprises and writes the value that begins to operate corresponding to, and wherein upgrade one or more data structures and comprise and write an opening entry that comprises the sign of Transaction Identifier and at least one memory devices, begin operation to carry out this.
46. as the method for claim 42, at least some contents that wherein write in data and the one or more data structure comprise the value that writes corresponding to a submit operation, and wherein upgrade one or more data structures and comprise:
The sign of the open state of affairs has been indicated in removing;
Write a submission record that comprises the sign of Transaction Identifier and at least one memory devices, to carry out this submit operation; And
Upgrade a transaction status daily record.
47. as the method for claim 42, at least some contents that wherein write in data and the one or more data structure comprise the value that writes corresponding to a terminating operation, and wherein upgrade one or more data structures and comprise:
The sign of the open state of affairs has been indicated in removing;
Write the termination record that comprises Transaction Identifier and at least one memory devices sign, to carry out this terminating operation; And
Upgrade a transaction status daily record.
48., comprise the state of the one or more data structures of scanning with the specified data affairs as the method for claim 36.
49., comprise and carry out a check point operation that comprises the following steps as the method for claim 36:
Pin one or more data structures;
The content of one or more data structures is write the check point position of at least one memory devices; And
In response to content being write at least one memory devices, and change at least some contents of one or more data structures.
50., comprise and carry out a recovery operation that comprises the following steps as the method for claim 49:
From at least one memory devices, read the content of one or more data structures; And
In response to foregoing, and upgrade one or more data structures at least one volatile memory.
51., wherein carry out recovery operation and comprise the content that reads since carrying out all one or more data structures that write since the check point operation as the method for claim 50.
52. method as claim 50, wherein carry out recovery operation and comprise the content that reads since carrying out all one or more data structures that write since the check point operation, the used time is not more than the time that all one or more data structures spent that has write since the check point operation since carrying out that writes.
53., wherein carry out recovery operation and comprise that the state and the feasible one or more clients that open data transactions that determine an open data transactions can inquire about this state and determine whether to submit to or stop this opening data transactions as the method for claim 50.
54., comprise and carry out the transaction coordination operation that comprises the following steps as the method for claim 36:
Select a register at least one memory devices as the data transactions state;
A beginning transaction operation is submitted to register;
The response of data transactions state that has been an indication that comes from least one memory devices, and wait for a schedule time;
Based on this response, one or more soft write operations are submitted at least one memory devices; And
Based on this response, determine whether to submit to or stop transaction operation to submit or stop this affairs to by issuing one.
55., comprise and carry out the affairs complete operation that comprises the following steps as the method for claim 36:
From at least one memory devices that contains the content of having indicated the data transactions state, read consecutive data block first group; And
In response to first group content of this consecutive data block, second group of consecutive data block write at least one memory devices and with one group of one or more data structure storage at least one volatile memory.
56. as the method for claim 31, wherein data transactions comprises a plurality of concurrent data transactions.
57. as the method for claim 31, wherein at least one memory devices comprises the disk by a server admin.
58. as the method for claim 31, wherein at least one memory devices comprises a plurality of memory devices.
59. be used for the equipment of electronic data storage, it is characterized in that described equipment comprises:
A memory devices is used for receiving consecutive data block to store wherein physical address separately into; And
Control circuit is configured when write store equipment, with consecutive data block and at least some data blocks and indicated in the consecutive data block one or more parameter write store equipment of the physical address of data block afterwards.
60. a method that is used for electronic data storage comprises:
Provide consecutive data block to store the physical address separately of a memory devices into;
For at least some data blocks in the consecutive data block each has been determined to indicate in the consecutive data block one or more parameters of the physical location of data block afterwards; And
Storage consecutive data block and one or more parameter in memory devices.
CNB011258179A 2000-08-30 2001-08-29 Apparatus and method for supporting data transactions Expired - Fee Related CN1190741C (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US09/651,910 US6668304B1 (en) 2000-01-18 2000-08-30 Transaction support on logical disks
US09/651910 2000-08-30

Publications (2)

Publication Number Publication Date
CN1365056A CN1365056A (en) 2002-08-21
CN1190741C true CN1190741C (en) 2005-02-23

Family

ID=24614740

Family Applications (1)

Application Number Title Priority Date Filing Date
CNB011258179A Expired - Fee Related CN1190741C (en) 2000-08-30 2001-08-29 Apparatus and method for supporting data transactions

Country Status (5)

Country Link
JP (1) JP2002149454A (en)
KR (1) KR100420549B1 (en)
CN (1) CN1190741C (en)
SG (1) SG99941A1 (en)
TW (1) TW522303B (en)

Families Citing this family (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100898228B1 (en) * 2002-09-16 2009-05-18 주식회사 케이티 Tablespace automatic allocation device and thereof method
US8495131B2 (en) 2002-10-08 2013-07-23 International Business Machines Corporation Method, system, and program for managing locks enabling access to a shared resource
US7496574B2 (en) 2003-05-01 2009-02-24 International Business Machines Corporation Managing locks and transactions
US11734393B2 (en) 2004-09-20 2023-08-22 Warner Bros. Entertainment Inc. Content distribution with renewable content protection
CN101542457B (en) * 2005-04-29 2015-11-25 微软技术许可有限责任公司 Transaction transforms
US7392335B2 (en) * 2006-02-10 2008-06-24 Oracle International Corporation Anticipatory changes to resources managed by locks
JP4740766B2 (en) * 2006-02-27 2011-08-03 富士通株式会社 Data receiving apparatus, data transmitting / receiving system, data transmitting / receiving system control method, and data receiving apparatus control program
GB0616068D0 (en) * 2006-08-12 2006-09-20 Ibm Method,Apparatus And Computer Program For Transaction Recovery
KR101132389B1 (en) * 2007-04-09 2012-04-03 엘지엔시스(주) Apparatus and method of structuralizing checkpoint memory based dispersion data structure
CA2714680C (en) 2008-02-11 2016-06-07 Kenneth Goeller Simple non-autonomous peering environment watermarking, authentication and binding
CN102999434A (en) * 2011-09-15 2013-03-27 阿里巴巴集团控股有限公司 Memory management method and device
US9348642B2 (en) 2012-06-15 2016-05-24 International Business Machines Corporation Transaction begin/end instructions
US9448796B2 (en) 2012-06-15 2016-09-20 International Business Machines Corporation Restricted instructions in transactional execution
US10437602B2 (en) 2012-06-15 2019-10-08 International Business Machines Corporation Program interruption filtering in transactional execution
US9740549B2 (en) 2012-06-15 2017-08-22 International Business Machines Corporation Facilitating transaction completion subsequent to repeated aborts of the transaction
US9384004B2 (en) 2012-06-15 2016-07-05 International Business Machines Corporation Randomized testing within transactional execution
US20130339680A1 (en) 2012-06-15 2013-12-19 International Business Machines Corporation Nontransactional store instruction
US10509658B2 (en) * 2012-07-06 2019-12-17 Nvidia Corporation System, method, and computer program product for simultaneously determining settings for a plurality of parameter variations
CN103578207B (en) * 2012-07-20 2015-12-02 中国移动通信集团公司 A kind of data processing method, device, system and relevant device
CN103235747B (en) * 2013-04-24 2016-12-28 曙光信息产业(北京)有限公司 The restoration methods of metadata and system
CN103605583B (en) * 2013-07-08 2017-03-15 宇龙计算机通信科技(深圳)有限公司 File access pattern method and system
KR102593362B1 (en) * 2016-04-27 2023-10-25 에스케이하이닉스 주식회사 Memory system and operating method of memory system
KR102694946B1 (en) * 2018-10-01 2024-08-14 에스케이하이닉스 주식회사 Memory system, operation method thereof and database system including the memory system

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH02310665A (en) * 1989-05-25 1990-12-26 Toshiba Corp Data restoration method for distributed transaction processing system
US5481694A (en) * 1991-09-26 1996-01-02 Hewlett-Packard Company High performance multiple-unit electronic data storage system with checkpoint logs for rapid failure recovery
JPH0619761A (en) * 1992-07-06 1994-01-28 Fujitsu Ltd File management system
JPH07175700A (en) * 1993-12-20 1995-07-14 Fujitsu Ltd Database management system
US5630047A (en) * 1995-09-12 1997-05-13 Lucent Technologies Inc. Method for software error recovery using consistent global checkpoints
JPH0991183A (en) * 1995-09-27 1997-04-04 Toshiba Corp Data base recovery device
US5799305A (en) * 1995-11-02 1998-08-25 Informix Software, Inc. Method of commitment in a distributed database transaction
JPH1153235A (en) * 1997-08-08 1999-02-26 Toshiba Corp Data updating method of disk storage device and disk storage control system
US6202067B1 (en) * 1998-04-07 2001-03-13 Lucent Technologies, Inc. Method and apparatus for correct and complete transactions in a fault tolerant distributed database system
JP2000163294A (en) * 1998-11-30 2000-06-16 Nec Corp Method and device for managing database and machine- readable recording medium with program recorded thereon

Also Published As

Publication number Publication date
CN1365056A (en) 2002-08-21
SG99941A1 (en) 2003-11-27
TW522303B (en) 2003-03-01
JP2002149454A (en) 2002-05-24
KR100420549B1 (en) 2004-03-02
KR20020018158A (en) 2002-03-07

Similar Documents

Publication Publication Date Title
CN1190741C (en) Apparatus and method for supporting data transactions
CN1308851C (en) Data processing system including storage systems
CN1271630C (en) Method of manufacturing information recording medium, information recording method, and information recording apparatus
CN1270270C (en) Proximity communication system, proximity communication method, data managing apparatus, data managing method, recording medium, and computer program
CN1146786C (en) System and method for automatically modifying database access methods to insert database object handling instructions
CN1752947A (en) Memory storage, storage controlling method, and computer product
CN1437110A (en) Arrangement and method for break-point setting
CN101042676A (en) Storage system, storage extent release method and storage apparatus
CN1041839A (en) The object management facility that includes the snapshot facility that between two objects, carries out the data transmission
CN1724980A (en) Storage system
CN1779660A (en) Methods for duplicating among three units asynchronously
CN1912926A (en) Method for stopping multiple programs in image processing unit and image processing unit
CN1916962A (en) Method and device storing and back-storing state context in graphic processing unit
CN1916961A (en) Interruptible graphic processing unit and its control method
CN1680942A (en) Document group analyzing apparatus, a document group analyzing method, a document group analyzing system
CN101075203A (en) Storage system and method of producing recovery volume
CN1779650A (en) Method and system for automatically managing the state of replicated data of a computing environment
CN1882906A (en) System and method for maintaining temporal data in data storage
CN1648867A (en) Data processing system
CN1174319C (en) Data structure managing device, data structure managing system, data structure managing method, and computer reconded medium for data structure managing program
CN1498367A (en) Information processing device, momery management device, memory management method and information processing method
CN1684029A (en) Storage system
CN101067955A (en) Content list display method, content list display apparatus, content selecting and processing method, and content selecting and processing apparatus
CN1218223A (en) Method of executing mobile objects and recording medium storing mobile objects
CN1610879A (en) Method of optimizing the performance of a computer and a main non-volatile memory unit optimised using specific-use sections

Legal Events

Date Code Title Description
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C06 Publication
PB01 Publication
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20050223

Termination date: 20160829