WO2017035759A1 - Method and system of rapid data backup on a solid state drive upon abnormal power off - Google Patents

Abstract

A method of rapid data backup on a solid state drive upon abnormal power off is applicable to the field of solid state storage, said method comprising: step A: monitoring whether abnormal power off occurs; step B: if abnormal power off is monitored, calculating whether the energy released by a backup capacitor meets the requirement for data backup; step C: if it is determined that the energy released by the backup capacitor meets the requirement for data backup, determining an idle channel as the write channel for data backup; step D: performing data backup by means of the idle channel. The method, according to the discharge characteristics of a backup capacitor, uses dynamic channel selections to fully utilize all current available channels to backup data, thereby improving the speed of data backup; further, in order to ensure the reliability of the backup data, the method of only programming LSB pages is used during backup, improving the backup efficiency, reducing the bit error rate of the backup data, and avoiding low page data corruption.

Description

 Method and system for fast backup of solid state hard disk data during abnormal power failure Technical field

 The invention belongs to the field of solid-state storage, and in particular relates to a method and a system for quickly backing up data of a solid-state hard disk based on an abnormal power-off of a capacitor.

Background technique

NAND As a non-volatile memory device, Flash has excellent features such as small size, fast access speed, low power and shock resistance. Therefore, solid state hard disk based on NAND Flash technology (Solid State) Drives, SSD) is currently used in military and residential storage. There are two main types of NAND flash: SLC NAND flash (Single-Level-Cell) NAND flash, single-layer NAND flash) and MLC NAND flash (Multi-Level-Cell NAND flash) Flash), MLC NAND flash is more widely used due to its high storage density and low price. MLC NAND Flash uses a technique called "multi-page structure", that is, a physical page is divided into two logical pages, which are LSB (Least Significant Bit) pages and MSB (Most). Significant Bit) page. Due to the MLC NAND When programming a physical page of flash, the LSB page must be programmed first, and then the MSB page is programmed. Therefore, the LSB page corresponds to a lower reference voltage. Therefore, the LSB page has a faster programming speed, and when only the LSB page is used, the MLC NAND flash can achieve a lower bit error rate similar to SLC NAND flash of the same process.

In the actual use of SSDs, unexpected power failures occur. Due to its unpredictability, accidental power loss can cause temporary data loss of SSDs. These temporary data may be address mapping tables or file system key data. Data loss, the SSD itself and the user system are at risk of crashing. Therefore, large capacitors have been widely used in solid state drives as a backup power source for accidental power failure. The typical storage system structure of a solid-state hard disk based on a large capacitor is shown in Figure 1. The solid-state hard disk power supply part includes a main power supply and a backup capacitor. The main power supply is generally provided by a main interface (such as USB, SATA, etc.), and is a solid-state storage system in normal operation. Provide power and charge the backup capacitor. One or more large capacitors store energy during normal operation and are used as a backup power source when the system is unexpectedly powered down. Since the voltage is unstable during the discharge process, the voltage regulator must be used to convert the capacitor voltage to provide a stable supply voltage for the solid-state storage system. The voltage regulator is generally a Zener diode, DC/DC, and the like. Solid state storage system by NAND A flash array and an SSD (Solid State Drive) controller. In order to increase the capacity of the solid-state storage system and reduce the package of the flash controller, multi-chip NAND is generally used. The flash chip is composed of multiple channels, which are then connected to the flash controller. As shown in FIG. 1 , the SSD controller is composed of a processor, a cache controller, a RAM, and a flash controller. The SSD controller mainly receives access requests from the main interface, performs address mapping operations on the requests, and finally controls the Flash. The controller completes these requests.

Large capacitor capacity is generally limited, and its capacity will gradually decrease with time. Therefore, it is necessary to enhance the backup efficiency and reliability of solid-state hard disk storage systems based on large capacitance. In order to improve the efficiency and reliability of backup data, the following issues need to be resolved:

(1) Due to backup capacitor life and capacity limitations, backup data needs to be backed up to NAND as quickly as possible In the flash. Since the system's unexpected power loss is generally unpredictable, the system may be able to handle some time-consuming tasks, such as NAND, when the system is powered down. Flash erase operation, therefore, how to achieve safe and fast data backup by effectively scheduling the use of NAND flash is an urgent problem to be solved when these situations occur;

(2) Since the backup capacitor is unstable during the discharge process, although the voltage regulator is generally regulated, the discharge process is unpredictable. Therefore, if the accidental power failure occurs, only the data is increased. Backup efficiency may cause some data reliability problems;

(3) If in MLC NAND When the MSB page of the flash is programmed to be powered off unexpectedly, low page data corruption occurs (lower page corruption) Phenomenon), that is, when the power is turned on again, the error rate of the already programmed LSB page data will rise significantly, which further jeopardizes the security of the backup data.

technical problem

The technical problem to be solved by the present invention is to provide a method and a system for quickly backing up data of a solid state hard disk during abnormal power failure, aiming at solving the problem that the prior art cannot solve the problem of safe, fast, accurate and effective data backup when the solid state hard disk is abnormally powered off. The problem.

Technical solution

The present invention is implemented in such a manner that a method for quickly backing up data of a solid state hard disk during abnormal power failure includes the following steps:

Step A, monitoring whether an abnormal power failure occurs;

Step B: If an abnormal power failure is detected, calculate whether the release energy of the backup capacitor satisfies the requirement of data backup;

Step C: If it is determined that the release energy of the backup capacitor can meet the data backup requirement, determine an idle channel, and use the idle channel as a write channel for data backup;

In step D, data backup is performed through the idle channel.

Beneficial effect

Compared with the prior art, the present invention has the beneficial effects that the present invention adopts a dynamic channel selection method according to the discharge characteristic of the backup capacitor, and fully utilizes all available channels to back up data, thereby improving the data backup speed, and further, To ensure the reliability of the backup data, the method of programming only the LSB page is used for backup, which improves the backup efficiency, reduces the error rate of the backup data, and avoids the destruction of low page data.

DRAWINGS

FIG. 1 is a schematic structural diagram of a typical storage system of a solid capacitor based on a large capacitor provided by the prior art.

FIG. 2 is a schematic structural diagram of an intelligent backup system according to an embodiment of the present invention.

FIG. 3 is a schematic diagram of an execution process of a data backup and data recovery task according to an embodiment of the present invention.

FIG. 4 is a flowchart of a method for quickly backing up data of a solid state hard disk during abnormal power failure according to an embodiment of the present invention.

FIG. 5 is a schematic flowchart of data backup and data recovery according to an embodiment of the present invention.

FIG. 6 is a schematic structural diagram of a solid state hard disk data fast backup system during abnormal power failure according to an embodiment of the present invention.

Embodiments of the invention

The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

In the present invention, the intelligent backup task is performed in the RAM and NAND when the power is abnormally powered off. A layer of software between Flash controllers. The structure of the solid-state storage system with intelligent backup function is shown in Figure 2. The host and the storage system are connected through the main interface. The main interface is also the main power supply of the storage system and the charging power supply of the backup power module during normal operation. All requests to the host file system are made by FTL (Flash Translation Layer, Flash translation layer) is done by controlling the NAND Flash controller. FTL mainly completes file system logical address to NAND Flash physical address mapping, garbage collection and wear leveling. In order to increase the system I/O rate and reduce system latency, all or part of the FTL mapping table, current pending requests, and file system key data are stored in RAM for quick access and update.

In the event of an abnormal power loss, the backup capacitor acts as a temporary power source to power the storage system for data backup. At the same time, the system generates a hardware interrupt, and the data backup task will be executed as an interrupt service routine. Figure 3 depicts the execution order and priority of data backup tasks and data recovery tasks. The purpose of data backup is to write data from RAM to NAND Flash arrays to avoid data loss. When the main power is restored, the data recovery task is first executed because of its higher priority, and the backup data in the NAND flash array is written to the RAM.

In order to speed up the backup process while ensuring the reliability of backup data, smart backup in NAND A fixed physical partition is separately created in the Flash array to store backup data. In order to utilize the parallelism of the NAND Flash array, the partition is evenly distributed in NAND. In each physical channel of the Flash array. Because NAND Flash is a "non-instant" updated storage device, namely NAND Each physical block of flash must first be erased before writing data, so NAND Flash arrays require garbage collection operations to reclaim "dirty blocks" of written data. In the smart backup method, all physical blocks of the backup data partition will be erased together during data recovery to prepare for the next backup data write. In order to obtain the real-time status of the backup capacitor to determine the number of channels for the next write parallel operation, the intelligent backup also needs to obtain the real-time voltage of the backup capacitor. The real-time voltage of the capacitor can be realized by a hardware acquisition circuit, or by calculating a model for a large capacitor and a system.

Based on the above principle, the present invention provides a method for protecting data of a solid state hard disk during abnormal power failure as shown in FIG. 4, and the steps include:

S1, when abnormal power failure is detected, the system begins to enter the data backup process;

S2. Calculate whether the release energy of the backup capacitor satisfies the requirement of data backup. In this step, the release energy of the backup capacitor is calculated by the voltage of the backup capacitor and its capacitance value, and the release energy is compared with the backup energy required to back up all the data, if the release energy is less than The backup energy ends the backup and sends a signal to the system to indicate that the backup fails due to insufficient release of energy; if the release energy is greater than the backup energy, the following steps are continued.

S3: Calculate the number of channels that can work at the same time according to the current backup capacitor voltage, and obtain the corresponding channel accordingly.

S4, detecting the current state in the channel determined in step S3. In this step, since some channels of the SSD are likely to be performing some operations during an unexpected power failure, they cannot be used, so the status of the channel capable of data writing is first detected for data backup.

S5. Select an idle channel according to the detection result in step S4. In steps S3 to S5, in order to ensure the reliability of the write data and avoid the situation that multiple channels work at the same time, thereby causing the backup capacitor energy to be exhausted, before selecting the idle channel, the current capacitor must be calculated according to the voltage and capacity of the backup capacitor at this time. The number of channels that can support the write operation can be supported at the same time, and then one of the channels is selected for writing. Moreover, because in the case of accidental power failure, the channel for data backup that can be used can be written in parallel, so that all channels can simultaneously perform data backup, but the SSD controller can only operate on one channel at a time, so Only one channel can be selected for the idle channel to write.

S6. Determine whether the current physical page is the first page of the current physical block. In this step, once an available free channel is selected for data backup write operation, it is necessary to determine the current physical page of the data to be written (PPN). Current) Whether it is the first page of the current physical block (PBN current).

S7. If it is determined in step S6 that the current physical page is the first page of the current physical block, the current physical page is read to obtain a bad block flag.

S8. Determine, according to the bad block flag obtained by the bad block flag, whether the current physical block is a bad block.

S9. If it is determined that the current physical block is a bad block, the next physical block is directly read, and the process returns to step S8. In the above steps S6 to S9, the bad block flag is recorded to a fixed position (ie, a bad block flag bit) of each physical block of the backup partition, and when the bad block determination is performed, the bad block mark of the first page can be read. It is judged whether the current fast PBNcurrent is a bad block. If the current physical block PBNcurrent is a bad block, the next physical block is directly read, and the bad block judgment is performed again until a non-bad block is found for data writing.

S10. If it is determined in step S6 that the current physical page is not the first page of the current physical block, it is determined whether the current physical page is an LSB page. In this step, since the data is directly written into the LSB page, it is necessary to judge whether the current physical page is an LSB page.

S11. If it is determined that the current physical page PPNcurrent is not an LSB page, the next physical page is directly read until the LSB page is found.

S12. If the current physical page PPNcurrent is an LSB page, write the backup data of a physical page size into the data area of the current physical page PPNcurrent, and write the corresponding RAM address into the OOB; if the current physics determined in step S6 When the page PPNcurrent is the first page, the non-bad block flag needs to be written into the OOB.

S13. Determine whether all the data is backed up. If the data is not completely backed up, return to step S3; if it is determined that all the data is backed up, the backup process is ended.

In the above method, the intelligent backup first calculates the release energy of the backup capacitor by the backup capacitor voltage and its capacitance value, and compares it with the energy required to back up all the data. If the backup capacitor energy cannot back up all the backup data, the backup process Failure, and send a signal to the system; otherwise, the smart backup will use the dynamic channel selection method, that is, to detect the current state of all channels, and select one channel in all idle channels for the spare data write operation. Because intelligent backup needs to write multiple channels of SSD in parallel to improve the writing speed, in order to ensure the reliability of writing data and avoid the exhaustion of multiple channels simultaneously working backup capacitors, before selecting the idle channel, The intelligent backup must calculate the maximum number of simultaneous channels that can be supported at this time based on the voltage and capacity of the backup capacitor at this time, and then select one of the channels to perform the write operation. Once an available free channel is selected, if the current physical page (PPNcurrent) to be written is the first page of the current physical block (PBNcurrent), the bad block is judged first. Smart Backup logs bad block marks to a fixed location on the first page of each physical block of the backup partition. Therefore, as shown in FIG. 4, when the bad block determination is performed, the bad block flag of the first page is read to determine whether the PBNcurrent is a bad block. If the PBNcurrent is a bad block, the physical block is directly taken off, and once again. Bad blocks are judged until a non-bad block is found for data writing. When a non-bad block is selected, the smart backup writes the backup data to the LSB page of PBNcurrent, and writes the corresponding memory address into the OOB (free area). If PPNcurrent is the first page, the non-bad block is marked. It also needs to be written to OOB at the same time. In this way, the intelligent backup no longer needs to maintain a mapping table between a static physical address and a memory address, thereby reducing the problem of channel congestion caused by the static mapping table.

An example of intelligent backup is shown in Figure 5. In this example, assume that one physical page can hold four units of backup data. The backup partition has four physical blocks distributed in four channels. When an accidental power failure occurs, the intelligent backup writes the data in the memory to the backup area. Since channel 3 is currently busy, the intelligent backup writes the backup data in parallel to the LSB physical page in the channel 0 to channel 2 physical block. At the same time, the corresponding memory address is written into the OOB of the corresponding physical page. When the primary power is restored, Smart Backup needs to read data from the backup partition into RAM to recover the data. As shown in FIG. 5, the intelligent backup reads the backup partition in parallel. Since the physical page stores the backup and its corresponding memory address, the smart backup only needs to write the read backup partition data to the corresponding RAM address. When all data recovery is complete, Smart Backup will erase all physical blocks of the backup partition for the next backup operation.

FIG. 6 is a data protection system for a solid state hard disk during abnormal power failure according to an embodiment of the present invention, including:

The detecting unit 1 is configured to monitor whether an abnormal power failure occurs;

The calculating unit 2 is configured to calculate whether the release energy of the backup capacitor meets the requirement of data backup;

The channel selection unit 3 is configured to determine a free channel when the release energy of the backup capacitor can meet the data backup requirement, and use the idle channel as the data backup backup channel;

The backup unit 4 is configured to perform data backup through the idle channel.

Further, the calculation unit 2 includes:

The first calculating module 21 is configured to calculate a release energy of the backup capacitor by using a voltage and a capacitance value of the backup capacitor;

The comparison module 22 is configured to compare the release energy of the backup capacitor with the backup energy required to back up all the data;

The energy judging module 23 is configured to end the backup when the release energy is less than the backup energy, and send an end signal to the system; and when the release energy is greater than the backup energy, generate a A channel selection command is sent to the channel selection unit.

Further, the channel selection unit 3 includes:

a second calculation module 31, configured to calculate, according to the release energy, a number of channels that can simultaneously perform data backup, and obtain a corresponding channel according to the number of channels;

a state detecting module 32, configured to detect a current state of the acquired channel;

The channel selection module 33 is configured to select an idle channel according to the detection result of the state detection module 32; and use the idle channel as the write channel of the data backup.

Further, the backup unit 4 includes:

The page determining module 42 is configured to determine whether the current physical page of the data to be written is an LSB page;

The page search module 43 is configured to: when determining that the current physical page is not an LSB page, look down until an LSB page is found;

The backup module 44 is configured to determine that the current physical page is an LSB page, write data of a physical page size to be backed up into a data area of the current physical page, and write the corresponding RAM address into the OOB;

The backup determining module 45 is configured to confirm whether all the data is backed up, and if the data is all backed up, the data backup is ended;

If the data is not all backed up, it returns to calculation unit 2.

Further, the backup unit 4 further includes a bad block determining module 41, specifically for:

First, determining whether the current physical page is the first page of the current physical block;

Secondly, if it is determined that the current physical page is the first page of the current physical block, the current physical page is read to obtain the bad block flag bit;

Again, determining whether the current physical block is a bad block; if the current physical block is not a bad block, performing data backup through the idle channel;

Finally, if the current physical block is a bad block, the next physical block is searched for and the bad block judgment is performed again;

The backup module 44 is specifically configured to: when determining that the current physical page is an LSB page, write the backup data of a physical page size into the data area of the current physical page, and write the corresponding RAM address into the OOB; The page is the first page, you also need to write non-bad block tags into OOB.

Compared with the traditional accidental power-down backup method based on large-capacity solid-state storage system, the advantages of intelligent backup are mainly reflected in the following two points:

(1) Data backup time is greatly reduced. Intelligent backup uses dynamic channel selection to avoid write waits when certain channels are used, thus improving backup and speed. At the same time, intelligent backup writes data to LSB pages with faster programming speed. Improved I/O performance during backup;

(2) Guarantee the reliability of the backup process and backup data. During the data backup process, each time the backup data is written, the smart backup dynamically calculates the number of channels that perform the write operation in parallel through the voltage value of the current backup capacitor, which avoids the capacitor power shortage that may occur during backup. Problem; Smart Backup writes all backup data to the LSB page with very low error rate, ensuring the reliability of the backup data.

The above is only the preferred embodiment of the present invention, and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection of the present invention. Within the scope.

Claims (10)

1.  A method for quickly backing up data of a solid state hard disk during abnormal power failure, characterized in that the method comprises the following steps:

   Step A, monitoring whether an abnormal power failure occurs;

   Step B: If an abnormal power failure is detected, calculate whether the release energy of the backup capacitor satisfies the requirement of data backup;

   Step C: If it is determined that the release energy of the backup capacitor can meet the data backup requirement, determine an idle channel, and use the idle channel as a write channel for data backup;

   In step D, data backup is performed through the idle channel.
2. The method of claim 1 wherein step B comprises:

   Step B1, calculating a release energy of the backup capacitor by using a voltage and a capacitance value of the backup capacitor;

   Step B2, comparing the release energy of the backup capacitor with the backup energy required to back up all the data;

   Step B3, if the release energy is less than the backup energy, ending the backup and sending an end signal to the system;

   In step B4, if the release energy is greater than the backup energy, step C is performed.
3. The method of claim 1 wherein step C comprises:

   Step C1: Calculate the number of channels that can simultaneously perform data backup according to the release energy, and obtain a corresponding channel according to the number of channels;

   Step C2, detecting a current state of the acquired channel;

   In step C3, an idle channel is selected according to the detection result of step C2; and the idle channel is used as a write channel for the data backup.
4. The method of claim 1 wherein step D comprises:

   Step D1: determining whether the current physical page of the data to be written in the selected idle channel is an LSB page;

   In step D2, if it is determined that the current physical page is not an LSB page, the search is performed until an LSB page is found;

   Step D3, if it is determined that the current physical page is an LSB page, the data to be backed up by a physical page size is written into the data area of the current physical page, and the corresponding RAM address is written into the OOB;

   In step D4, it is confirmed whether all the data is backed up, and if all the data is backed up, the data backup is ended;

   If the data is not all backed up, return to step C.
5. The method according to claim 4, wherein in step D, performing bad data determination before performing data backup;

   The specific steps for bad block judgment include:

   Step E1: determining whether the current physical page is the first page of the current physical block;

   Step E2: If it is determined that the current physical page is the first page of the current physical block, reading the current physical page to obtain the bad block flag bit;

   Step E3: determining, according to the bad block flag obtained by the bad block flag, whether the current physical block is a bad block; if the current physical block is not a bad block, performing data backup through the idle channel;

   Step E4, if the current physical block is a bad block, then look for the next physical block and return to step E3;

   Specifically, in step D3, the following includes: If the current physical page is an LSB page, the backup data of a physical page size is written into the data area of the current physical page, and the corresponding RAM address is written into the OOB; if the current physical page is the first page, it will be non-bad. The block tag is written to the OOB.
6. A system for quickly backing up data of a solid state hard disk during abnormal power failure, characterized in that the system comprises:

   a detecting unit for monitoring whether an abnormal power failure occurs;

   The calculation unit is configured to monitor whether an abnormal power failure occurs, and calculate whether the release energy of the backup capacitor satisfies the requirement of data backup;

   The channel selection unit is configured to determine a free channel when the release energy of the backup capacitor can meet the data backup requirement, and use the idle channel as the data backup backup channel;

   a backup unit, configured to perform data backup through the idle channel.
7. The system of claim 6 wherein said computing unit comprises:

   a first calculating module, configured to calculate a release energy of the backup capacitor by using a voltage and a capacitance value of the backup capacitor;

   Comparison module for comparing the release energy of the backup capacitor with the backup energy required to back up all data;

   An energy judging module, configured to end the backup when the release energy is less than the backup energy, and send an end signal to the system; and configured to generate a channel when the release energy is greater than the backup energy Select an instruction to the channel selection unit.
8. The system of claim 6 wherein said channel selection unit comprises:

   a second calculating module, configured to calculate, according to the release energy, a number of channels that can simultaneously perform data backup, and obtain a corresponding channel according to the number of channels;

   a state detecting module, configured to detect a current state of the acquired channel;

   The channel selection module is configured to select an idle channel according to the detection result of the state detection module; and use the idle channel as the write channel of the data backup.
9. The system of claim 6 wherein the backup unit comprises:

   a page determining module, configured to determine whether the current physical page of the data to be written in the selected idle channel is an LSB page;

   The page search module is configured to search down until an LSB page is found when determining that the current physical page is not an LSB page;

   The backup module is configured to determine that the current physical page is an LSB page, write data of a physical page size to be backed up into a data area of the current physical page, and write the corresponding RAM address into the OOB;

   The backup determination module is configured to confirm whether all the data is backed up, and if the data is completely backed up, the data backup is ended;

   If the data is not fully backed up, return to the calculation unit.
10. The system of claim 9, wherein the backup unit further comprises a bad block determination module, and the bad block determination module is specifically configured to:

    First, determining whether the current physical page is the first page of the current physical block;

    Secondly, if it is determined that the current physical page is the first page of the current physical block, the current physical page is read to obtain the bad block flag bit;

    Again, determining whether the current physical block is a bad block; if the current physical block is not a bad block, performing data backup through the idle channel;

    Finally, if the current physical block is a bad block, the next physical block is searched for and the bad block judgment is performed again;

    The backup module is specifically configured to: when determining that the current physical page is an LSB page, write the backup data of a physical page size into the data area of the current physical page, and write the corresponding RAM address into the OOB; The page is the first page, and the non-bad block flag is written into OOB.