CN102567147B - Hard disk array controller based on double hard disk redundancy - Google Patents

Abstract

The invention discloses a hard disk array controller based on double hard disk redundancy, which comprises a first data buffer area, a second data buffer area, a third data buffer area, a fourth data buffer area, a data playback verification module, a redundant data generation module, a damaged hard disk reconstruction module, a data gate, a hard disk control interface and a hard disk array. The storage and reading rate of the hard disk array controller can be still kept unchanged when an optional hard disk or two optional hard disks in the hard disk array are in failure, the hard disk array controller can still realize the functions of recording while playing back and playing back while reconstructing when some hard disks in the array are in failure, and safety of the hard disk array is greatly enhanced.

Description

A kind of hard disk array controller based on two hard disk redundancy

Technical field

The present invention relates to a kind of hard disk array controller based on two hard disk redundancy.

Background technology

In high speed hard-disk array data field of storage, along with improving constantly of monolithic disk read-write speed and capacity, the memory rate of hard disk array and memory bandwidth are also in continuous lifting, and two-forty, large bandwidth, Large Copacity have become hard disk array field of data storage mainstream development trend.But thing followed hard disk array problem of data safety also highlights, in many instances, people have even exceeded the concern to data storage rate to the concern of problem of data safety.Particularly in space industry, data store and problem of data safety particularly important, often perform human and material resources that a subtask spends and financial resources all quite huge, therefore the data message obtained in task process is also very precious, if appropriate for these information securities can not be preserved, loss will be difficult to estimate, and even can cause the failure of whole task.Redundant data dish is added in hard disk array, when there being hard disk to lose efficacy in array, inefficacy hard disc data can be recovered according to redundant information the Main Means becoming and improve hard disk array security, but once occur there is hard disk corruptions situation in array in existing commercial hard disk array stores solution, DISK to Image array readwrite performance will be greatly affected, and even cannot reach the minimum memory property requirement of executing the task.

Summary of the invention

Technology of the present invention is dealt with problems and is: for the deficiencies in the prior art, provides a kind of hard disk array controller based on two hard disk redundancy.Adopt the present invention still can be able to keep when any one piece or two pieces of hard disks lost efficacy in hard disk array storing and reading rate constant, and the function of limit writing edge playback, the reconstruction of playback limit, limit still can be realized when there being hard disk to lose efficacy in array, hard disk array security is simultaneously greatly improved.

Technical solution of the present invention is:

A kind of hard disk array controller based on two hard disk redundancy, comprise: data buffer one, data buffer two, data buffer three, data buffer four, data readback correction verification module, redundant data generation module, damage hard disk rebuild module, data strobe, hard disk controlling interface and hard disk array, wherein, when recording data, after a pair, data buffer input front end data cushion, output to redundant data generation module; Redundant data generation module generates check information according to the front end data of input, and outputs to data buffer three through data strobe; Data buffer three is to being stored in hard disk array by front end data and check information by hard disk controlling interface; When playback of data, data buffer two outputs to data readback correction verification module after the record data that hard disk array exports being cushioned; Record data without when damaging hard disk, are directly exported by data buffer four by data readback correction verification module in hard disk array.

Described redundant data generation module, by band, checking treatment is carried out to front end data and generate the checking data corresponding with each band, and the checking data of front end data and generation is stored into respectively in data disks in hard disk array and check disk by the order of packet in band;

Described data readback correction verification module, when there is degrading data dish in hard disk array, checking data corresponding with band in the packet and check disk stored in data disks working properly in hard disk array is read by band, obtained the packet stored in degrading data dish in a band by Recovery processing, then by the record order of redundant data generation module, the packet of record is exported; Described Recovery processing and checking treatment are contrary process;

Module rebuild by described damage hard disk, when there is degrading data dish in hard disk array, checking data corresponding with band in the packet and check disk stored in the data disks of normal work in hard disk array is read by band, obtained the packet stored in degradation dish in a band by Recovery processing, and the packet recovered is stored in the data disks that hard disk array newly adds.

Described redundant data generation module carries out checking treatment by following formula to front end data and generates checking data:

$P_k=D(k,0)\⊕D(k,1)\⊕D(k,2)...\⊕D(k,n-1)$

$Q_k=2^0\⊗D(k,0)\⊕2^1\⊗D(k,1)\⊕...\⊕2^{n-1}\⊗D(k,n-1)$

Wherein, P _kand Q _kfor the checking data according to a kth band; N is the hard disk sum in hard disk array.

Described data readback correction verification module or damage hard disk and rebuild module, by following formula, the data that Recovery processing recovers in degrading data dish are carried out to front end data:

$D_{(k,i)}={(2^i\⊕2^j)}^{-1}$

$\⊗\{2^0\⊗D_{(k,0)}\⊕2^1\⊗D_{(k,1)}\⊕......\⊕2^{i-1}\⊗D_{(k,i-1)}$

$\⊕2^{i+1}\⊗D_{(k,i+1)}\⊕......\⊕2^{j-1}\⊗D_{(k,j-1)}\⊕2^{j+1}\⊗D_{(k,j+1)}$

$\⊕......\⊕2^n\⊗D_{(k,n)}\⊕Q_{\left(k\right)}\⊕2^j$

$\⊗[D_{(k,0)}\⊕D_{(k,1)}\⊕......{\⊕D}_{(k,i-1)}\⊕D_{(k,i+1)}\⊕......\⊕D_{(k,j-1)}$

$\⊕D_{(k,j+1)}\⊕......\⊕D_{(k,n)}\⊕P_{\left(k\right)}\left]\right\}$

$D_{(k,j)}={(2^i\⊕2^j)}^{-1}$

$\⊗\{2^0\⊗D_{(k,0)}\⊕2^1\⊗D_{(k,1)}\⊕......\⊕2^{i-1}\⊗D_{(k,i-1)}$

$\⊕2^{i+1}\⊗D_{(k,i+1)}\⊕......\⊕2^{j-1}\⊗D_{(k,j-1)}\⊕2^{j+1}\⊗D_{(k,j+1)}$

$\⊕......\⊕2^n\⊗D_{(k,n)}\⊕Q_{\left(k\right)}\⊕2^i$

$\⊗[D_{(k,0)}\⊕D_{(k,1)}\⊕......\⊕D_{(k,i-1)}\⊕D_{(k,i+1)}\⊕......\⊕D_{(k,j-1)}$

$\⊕D_{(k,j+1)}\⊕......\⊕D_{(k,n)}\⊕P_{\left(k\right)}\left]\right\}$

Wherein k is band numbering, and n is hard disk serial number, and i, j are degradation hard disk serial number.

The present invention compared with prior art tool has the following advantages:

(1) the present invention is based on the hard disk array real-time calibration technology at a high speed of two hard disk redundancy, to the situation that hard disk array degradation uses, by data, Reconstruction of The Function and data real-time playback function are separated and are rebuild module and data readback correction verification module completes by damaging hard disk respectively afterwards, when damage hard disk reconstruction module mainly performs in hard disk array and there is degradation (inefficacy) data disks, the data of record are rebuild, the checking data of packet and the correspondence stored in other data disks can be utilized by the date restoring of degraded fail data dish out when system is in the non task period, data readback correction verification module mainly performs the data readback function when there being data disks to lose efficacy, when system is in the task period and has one piece or two blocks of data dishes to lose efficacy in array by Recovery processing by the data real-time recovery of fail data dish and playback to front end, therefore ensure that being recorded data still can carry out the storage of data and reading and not reduction of speed according to original speed when having one piece or two blocks of data dishes to lose efficacy,

(2) in the present invention, data readback correction verification module collects strip data when data readback, when losing efficacy without hard disk, effective data packets is exported to front end and shield checking data bag simultaneously, when there being data disks to lose efficacy, also can export effective data packets according to checking algorithm and shield checking data bag simultaneously, realize the complete playback of band valid data.

(3) the present invention make full use of FPGA device can the feature of concurrent operation, the real-time calibration technology at a high speed of the hard disk array based on two hard disk redundancy being applicable to FPGA realization is proposed targetedly, adopt the framework of Multi-modules parallel processing, this framework ensure that the real-time of data processing, date restoring when checking algorithm based on two hard disk redundancy can support that in array, maximum two pieces of hard disks lost efficacy simultaneously, relative monolithic hard disk redundancy array, the checking algorithm of two hard disk redundancy is from the reliability greatly enhancing system.

Accompanying drawing explanation

Fig. 1 is schematic diagram of the present invention;

Fig. 2 is hard disk array strip data storage organization schematic diagram of the present invention;

Fig. 3 is P check disk redundant data generating algorithm structural representation of the present invention;

Fig. 4 is Q check disk redundant data generating algorithm structural representation of the present invention;

Fig. 5 is that mould of the present invention takes advantage of modular algorithm structural representation;

The table of logarithm that Fig. 6 takes advantage of module to use for mould of the present invention;

The antilogarithm table that Fig. 7 takes advantage of module to use for mould of the present invention.

Embodiment

Just by reference to the accompanying drawings the present invention is described further below.

Be illustrated in figure 1 and the present invention includes: data buffer one, data buffer two, data buffer three, data buffer four, data readback correction verification module, redundant data generation module, damage hard disk rebuild module, data strobe, hard disk controlling interface and hard disk array

Front end data source, data buffer one, redundant data generation module, data strobe, data buffer three, hard disk controlling interface, hard disk array, according to data flow direction composition data record branch road.Wherein, front end data source is the Data entries needing to store, and exchanges data adopts packet switch framework receiving front-end packet.

Hard disk array, hard disk controlling interface, data buffer two, data readback correction verification module, data buffer four, plays back to front-end interface, by data flow composition data playback branch road.

Hard disk array, hard disk controlling interface, data buffer two, damage hard disk and rebuild module, data strobe, data buffer three, hard disk controlling interface, hard disk array rebuilds branch road by data flow composition data.

Four data buffers each can cushion the data capacity of a stripe size, in band, data packet number is hard disk number in hard disk array, and redundancy check data generates and data real-time recovery all operates in units of band.

Redundant data generation module adopts the data genaration algorithm based on two hard disk redundancy, raw data is sent in order to the back-end, and generation redundancy check information is inserted between raw data simultaneously, this module work speed is by hard disk array degradation (inefficacy) impact.Data check algorithm based on two hard disk redundancy adopts mould to take advantage of the checking treatment algorithm combined with XOR, fast parallel arithmetic can go out P check disk and Q check disk redundant data information.

The checking treatment algorithmic formula of checking data is generated according to following the present invention:

$P_k=D(k,0)\⊕D(k,1)\⊕D(k,2)...\⊕D(k,n-1)$

$Q_k=2^0\⊗D(k,0)\⊕2^1\⊗D(k,1)\⊕...\⊕2^{n-1}\⊗D(k,n-1)$

Wherein represent XOR; symbol represents modular multiplication; Modular multiplication priority is higher than XOR priority; 0 ~ 11 represents 12 data disks in hard disk array.

Then in Fig. 2, in band arrangement, the first row STRIP0 is example, and the checking data of generation is as follows respectively:

$P=D_{\left(\mathrm{0,0}\right)}\⊕D_{\left(\mathrm{0,1}\right)}\⊕......\⊕D_{\left(\mathrm{0,12}\right)}$

$Q=2^0\⊗D_{\left(\mathrm{0,0}\right)}\⊕2^1\⊗D_{\left(\mathrm{0,1}\right)}\⊕......\⊕2^{12}\⊗D_{\left(\mathrm{0,12}\right)}$

Data readback correction verification module adopts the data check algorithm based on two hard disk redundancy, when having one piece or two pieces of hard disk corruptions in hard disk array, according to the packet in residue health data dish, real-time for valid data can be recovered, and original valid data are played back to front-end interface, this module work speed is not by hard disk array degradation impacts.The Recovery processing algorithm of two hard disk redundancy is the algorithm for inversion of the checking treatment algorithm of two hard disk redundancy, fast parallel arithmetic can go out the data of corrupt data dish, as follows.

$D_{(k,i)}={(2^i\⊕2^j)}^{-1}$

$\⊗\{2^0\⊗D_{(k,0)}\⊕2^1\⊗D_{(k,1)}\⊕......\⊕2^{i-1}\⊗D_{(k,i-1)}$

$\⊕2^{i+1}\⊗D_{(k,i+1)}\⊕......\⊕2^{j-1}\⊗D_{(k,j-1)}\⊕2^{j+1}\⊗D_{(k,j+1)}$

$\⊕......\⊕2^n\⊗D_{(k,n)}\⊕Q_{\left(k\right)}\⊕2^i$

$\⊗[D_{(k,0)}\⊕D_{(k,1)}\⊕......\⊕D_{(k,i-1)}\⊕D_{(k,i+1)}\⊕......\⊕D_{(k,j-1)}$

$\⊕D_{(k,j+1)}\⊕......\⊕D_{(k,n)}\⊕P_{\left(k\right)}\left]\right\}$

$D_{(k,j)}={(2^i\⊕2^j)}^{-1}$

$\⊗\{2^0\⊗D_{(k,0)}\⊕2^1\⊗D_{(k,1)}\⊕......\⊕2^{i-1}\⊗D_{(k,i-1)}$

$\⊕2^{i+1}\⊗D_{(k,i+1)}\⊕......\⊕2^{j-1}\⊗D_{(k,j-1)}\⊕2^{j+1}\⊗D_{(k,j+1)}$

$\⊕......\⊕2^n\⊗D_{(k,n)}\⊕Q_{\left(k\right)}\⊕2^i$

$\⊗[D_{(k,0)}\⊕D_{(k,1)}\⊕......\⊕D_{(k,i-1)}\⊕D_{(k,i+1)}\⊕......\⊕D_{(k,j-1)}$

$\⊕D_{(k,j+1)}\⊕......\⊕D_{(k,n)}\⊕P_{\left(k\right)}\left]\right\}$

Wherein k is band numbering, and n is hard disk serial number, and i, j are degradation hard disk serial number.

Equally for the first row STRIP0 in band arrangement in Fig. 2, when having data disks 8 and number in array

When damaging according to dish 11 liang of blocks of data dishes, recover D simultaneously _(0,8)and D _(0,11)method as follows:

$D_{\left(\mathrm{0,8}\right)}={(2^8\⊕2^{11})}^{-1}$

$\⊗\{2^0\⊗D_{(0,0)}\⊕2^1\⊗D_{(0,1)}\⊕......\⊕2^7\⊗D_{(0,7)}$

$\⊕2^9\⊗D_{(0,9)}\⊕2^{10}\⊗D_{(0,10)}\⊕2^{12}\⊗D_{(0,12)}\⊕Q_{\left(0\right)}\⊕2^{11}$

$\⊗[D_{(0,0)}\⊕D_{(0,1)}\⊕......{\⊕D}_{(0,7)}\⊕D_{(0,9)}\⊕D_{\left(\mathrm{0,10}\right)}\⊕D_{(0,12)}$

$\⊕P_{\left(0\right)}\left]\right\}$

$D_{\left(\mathrm{0,12}\right)}={(2^8\⊕2^{11})}^{-1}$

$\⊗\{2^0\⊗D_{(0,0)}\⊕2^1\⊗D_{(0,1)}\⊕......\⊕2^7\⊗D_{(0,7)}$

$\⊕2^9\⊗D_{(0,9)}\⊕2^{10}\⊗D_{(0,10)}\⊕2^{12}\⊗D_{(0,12)}\⊕Q_{\left(0\right)}\⊕2^8$

$\⊗[D_{(0,0)}\⊕D_{(0,1)}\⊕......{\⊕D}_{(0,7)}\⊕D_{(0,9)}\⊕D_{\left(\mathrm{0,10}\right)}\⊕D_{(0,12)}$

$\⊕P_{\left(0\right)}\left]\right\}$

Wherein P ₍₀₎represent the checking data that P check disk stores, Q ₍₀₎represent the checking data that Q check disk stores, because two blocks of data hard disks damage the situation comprising monolithic data hard disk corruptions simultaneously, therefore the restoration methods of monolithic data dish is contained in above formula, and getting wherein any one expression formula can be restored.

Damage hard disk and rebuild the data check algorithm of module employing based on two hard disk redundancy, when having one piece or 2 blocks of data adjustment debit bad times in hard disk array, can according to the data in residue health data dish by corrupt data dish data real-time recovery out, and data after recovery are write in the new data dish of replacing, this module work speed is not by hard disk array degradation impacts.Recovery processing algorithmic formula is identical with the Recovery processing algorithmic formula of data readback correction verification module.

The effect of data strobe module makes gating to the data that redundant data generation module and damage hard disk reconstruction module export, when array is in logging mode, select the data of redundant data generation module output to rear module, when array is in hard disk reconstruct mode, selects to damage hard disk and rebuild the data of module output to rear module.

Hard disk controlling interface module completes the read-write operation function to hard disk array monolithic hard disk.Adopt parallel processing architecture, reading while write of polylith hard disk can be completed simultaneously.

Fig. 2 is hard disk array strip data storage organization schematic diagram of the present invention, the hard disk array that this schematic diagram forms for 14 pieces of hard disks, operate to improve data storage rate for realizing many hard disks simultaneously, data write in hard disk array with striped patterns, form a data strip at the packet of the synchronization write each blocks of data dish of hard disk array and check disk, in same data strip, comprise 12 effective data packets, 1 P check disk packet and 1 Q check disk packet.When carrying out data restore operation also in units of band by date restoring out.

According to above-mentioned checking treatment algorithmic formula, the present invention generates the checking treatment structure of redundant data in P check disk as shown in Figure 3.P check disk redundant data is the result of the mutual XOR of each effective data packets in band, its generation module adopts circulation XOR mode, be made up of an XOR gate and two twoport ram, first effective data packets is first stored in twoport ram A, read the first bag data stored in twoport ram A when second packet enters simultaneously, be stored in twoport ram B after XOR gate XOR, read data in twoport ram B when three guarantees data enter simultaneously, crossing after XOR gate carries out XOR is stored in twoport ram A, circulation like this is until the result that last packet obtains is P check disk redundant data.

In like manner, according to Recovery processing algorithm, Fig. 4 is the checking treatment structural representation that the present invention generates Q check disk redundant data, similar with P check disk generation method, Q check disk data generation module also adopts circulation XOR framework, unlike the step first will carrying out modular multiplication in data before XOR gate.

Fig. 5 is that mould of the present invention takes advantage of modular algorithm structural representation, and ordinary multiplications can produce data carry issue after data are multiplied, and the feature that mould is taken advantage of be two number be multiplied after can not produce data carry issue.It is patrol in China (Galois) territory at character used in proper names and in rendering some foreign names to do multiplication that mould is taken advantage of, and does mould and takes advantage of, be equivalent at GF (2 between 8bit bit wide data ⁸) do multiplication in territory, there is following features in this territory:

It is a number field, comprises 256 elements, namely 0,1,2 ..., 255}, if with 16 systems represent wherein element namely 00,01,02 ..., FF}.

It is a closed domain, and the result that territory interior element performs the four operations must drop in territory, is similar to the periodicity of trigonometric function.

All computings are all man-to-man mappings, i.e. the inverse operation of all computings obtains unique consequence.

As field element is converted into 2 systems, its each position can regard polynomial expression a as ₇x ⁷+ a ₆x ⁶+ a ₅x ⁵+ a ₄x ⁴+ a ₃x ³+ a ₂x ²+ a ₁x+a ₀coefficient, and polynomial expression X ⁸+ X ⁴+ X ³+ X ²+ 1 primitive polynomial being called this territory, the root of primitive polynomial is called primitive element, and each power of primitive element can generate all elements in territory.

Therefore the method that in this territory, two digital-to-analogues are taken advantage of is equal to take the logarithm respectively to two data and sues for peace, the process of summed result negate logarithm again.It asks the method for logarithm and inverse logarithm to adopt lookup table mode to realize, and Fig. 6 is 8bit data table of logarithm, and Fig. 7 is 8bit data antilogarithm table.Two tables are stored in two rom storeies, input data are read address as rom storer, obtain result and be checking result.

The unspecified part of the present invention belongs to general knowledge as well known to those skilled in the art.

Claims (2)

1. the hard disk array controller based on two hard disk redundancy, comprise: data buffer one, data buffer two, data buffer three, data buffer four, data readback correction verification module, redundant data generation module, damage hard disk rebuild module, data strobe, hard disk controlling interface and hard disk array, wherein, when recording data, after a pair, data buffer input front end data cushion, output to redundant data generation module; Redundant data generation module generates check information according to the front end data of input, and outputs to data buffer three through data strobe; Data buffer three is to being stored in hard disk array by front end data and check information by hard disk controlling interface; When playback of data, data buffer two outputs to data readback correction verification module after the record data that hard disk array exports being cushioned; Record data without when damaging hard disk, are directly exported by data buffer four by data readback correction verification module in hard disk array; It is characterized in that: following module realizes in FPGA;

Described redundant data generation module, by band, checking treatment is carried out to front end data and generate the checking data corresponding with each band, and the checking data of front end data and generation is stored into respectively in data disks in hard disk array and check disk by the order of packet in band;

Described data readback correction verification module, hard disk array exist any one piece or two pieces of hard disks lost efficacy time, checking data corresponding with band in the packet and check disk stored in data disks working properly in hard disk array is read by band, obtained the packet stored in degrading data dish in a band by Recovery processing, then by the record order of redundant data generation module, the packet of record is exported; Described Recovery processing and checking treatment are contrary process; Described data readback correction verification module is when system is in the task period and has one piece or two pieces of hard disks inefficacies in array, by two hard disk redundancy checking algorithm by the data real-time recovery of inefficacy hard disk and playback to front end, therefore ensure that array data still can carry out the storage of data and reading and not reduction of speed according to original speed when having one piece or two pieces of hard disks lost efficacy;

Module rebuild by described damage hard disk, when there is degrading data dish in hard disk array, checking data corresponding with band in the packet and check disk stored in the data disks of normal work in hard disk array is read by band, obtained the packet stored in degradation dish in a band by Recovery processing, and the packet recovered is stored in the data disks that hard disk array newly adds;

Described hard disk array controller makes full use of FPGA device can the feature of concurrent operation, the real-time calibration technology at a high speed of the hard disk array based on two hard disk redundancy being applicable to FPGA realization is proposed targetedly, adopt the framework of Multi-modules parallel processing, this framework ensure that the real-time of data processing.

2. a kind of hard disk array controller based on two hard disk redundancy as claimed in claim 1, is characterized in that: described redundant data generation module carries out checking treatment by following formula to front end data and generates checking data:

Wherein, P _kand Q _kfor the checking data according to a kth band; N is the hard disk sum in hard disk array;

Described data readback correction verification module or damage hard disk and rebuild module, by following formula, the data that Recovery processing recovers in degrading data dish are carried out to front end data:

Wherein k is band numbering, and n is hard disk serial number, and i, j are degradation hard disk serial number.