JP2006313407A - Dirty data processing method, dirty data processor and dirty data processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To secure accurate data assurance by shut-down processing after the occurrence of power failure, and to minimize the deterioration in the response performance of I/O processing since a power is restored until the charging of a battery is completed or when the battery device in normal operation breaks down and deteriorates. <P>SOLUTION: This dirty data processor is configured to monitor a residual battery capacity of a storage device, and to control the allowable quantity of dirty data in a cache memory of the storage device in accordance with the residual battery capacity, and to perform a write-through operation when the allowable quantity of the dirty data is a threshold or less, and to perform a write-back operation when the allowable quantity of the dirty data is a threshold or more. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to data guarantee of a RAID device and prevention of response performance degradation of I / O processing, particularly when the remaining battery capacity is reduced.

  Conventionally, when an unexpected power failure such as a power failure occurs in a RAID device, update data (hereinafter referred to as “dirty data”) that has not yet been written to the disk in the cache memory using a battery built in the device is temporarily stored. A process of writing to a disk for saving (hereinafter referred to as a “backup disk”) was performed.

  In the prior art, in order to increase the writing speed at normal times, for example, when writing to the cache memory, the writing destination disk (hereinafter referred to as “target disk”) is not written, but asynchronously with the I / O. A write back method is used in which data in the cache memory is written to the target disk and reflected on the target disk. And when the remaining battery capacity drops due to a battery failure or power failure, the time to write all dirty data to the backup disk is not guaranteed, so the cache data is written to the target disk in synchronization with I / O. The operation is switched to the write-through method. In the write-through method, there is a problem that the response performance of the I / O processing is deteriorated because the host computer receives a response only when writing to the target disk is completed.

  As a known example, for example, in Japanese Patent Laid-Open No. 7-146822, there is no need to minimize the system stoppage by making the backupable time such as a power outage by the uninterruptible power supply of the information processing system variable. The value corresponding to the backup possible time in the power failure power supply is stored in the register as an initial value, the value of the register is decreased by a timer in the event of a power failure, etc., and the data storage capacity of the disk cache memory while the OS monitors the register value A technique for transferring the data in the disk cache memory to the secondary storage device is disclosed in order to reduce the amount of data and make the data in the disk cache memory zero at the start of the shutdown process.

  In the above known example, the capacity of the disk cache memory is dynamically changed according to the time that can be backed up by the uninterruptible power supply, thereby starting the shutdown process such as a warning to the user or a stop process of the application program. It is a system that allows continuation of business as much as possible by delaying it to the point near the backup limit by the uninterruptible power supply, and uses the amount of disk cache memory corresponding to the time that can be backed up by the uninterruptible power supply. If an uninterruptible power supply fails during a power failure, the dirty data in the disk cache memory cannot be saved to the secondary storage device, and the I / O processing may become unstable during the power failure. is there. As another method, considering the failure of the uninterruptible power supply during a power outage, always estimate the time that can be backed up by the uninterruptible power supply with respect to the used capacity of the disk cache memory, and leave a margin. It can be considered, but it is wasteful.

In a RAID device, it is necessary to make data guarantee the top priority even when a power failure occurs or when the remaining battery capacity is abnormal during normal times, and it is necessary to minimize the degradation of response performance while performing I / O processing.
JP-A-7-146822

  The present invention ensures reliable data guarantee by the shutdown process after the occurrence of a power failure, and further performs I / O during the period from the time of power recovery until the completion of battery charging, or in the case of failure or deterioration of the battery device during normal operation. The purpose is to minimize the degradation of processing response performance.

  The invention according to claim 1 of the present invention is a step of monitoring a remaining battery capacity of a storage device, and a step of controlling an allowable amount of dirty data in a cache memory of the storage device according to the remaining battery capacity. A dirty-through operation when the allowable amount of dirty data is less than or equal to a threshold value, and a write-back operation when the allowable amount of dirty data is greater than the threshold value. Data processing method.

  According to the present invention, even when a power failure occurs in the storage device, the allowable amount of dirty data is changed depending on the remaining battery capacity so as to ensure the time for writing all dirty data to the backup disk. When the remaining battery capacity is a sufficient normal value, the cache memory is permitted to use the maximum allowable amount and is operated in write back.

  When the remaining battery capacity decreases from the normal value, a threshold is set, and the allowable amount of dirty data is reduced until the threshold is reached, and write-back operation is continued without shifting to write-through operation, and I / O processing To prevent response performance degradation. In this way, in order to guarantee the time to write all dirty data to the backup disk even if a power failure occurs, the amount of dirty data is changed depending on the remaining battery capacity, and the write-back operation shifts to the write-through operation. The process continues without doing so.

  When the remaining battery capacity further decreases and the remaining battery capacity reaches a threshold value, the write-back operation shifts to the write-through operation. In this way, even when the remaining battery capacity is reduced, the write-through operation is not performed regularly, and the operation shifts to the write-through operation when it falls below a certain threshold value. Even when the remaining battery capacity just before the power failure occurs is less than the normal value, the dirty data decreases depending on the remaining battery capacity, so the backup time at the time of a power failure is longer than the normal remaining battery capacity. The backup processing is ensured even when the remaining battery capacity is reduced. In the write-back operation when the remaining battery capacity does not reach the threshold value, the response of the I / O processing can be stably performed by appropriately selecting the threshold value. Further, when the remaining battery capacity further decreases and reaches the threshold value, the write-back operation is shifted to the write-through operation, so that a constant I / O response can be guaranteed.

  In this way, by controlling the allowable amount of dirty data according to the remaining battery capacity, and by setting a threshold value for the remaining battery capacity and using write-back operation and write-through operation separately, even when a power failure occurs, The evacuation to the backup disk is ensured and a certain I / O response can be guaranteed.

  Here, the detection of the occurrence of a power failure and the monitoring of the remaining battery capacity are performed by, for example, a known method.

  The invention according to claim 2 of the present invention is characterized in that the storage device performs control to make the dirty data allowable amount in the cache memory constant between the occurrence of a power failure and the start of shutdown. The dirty data processing method according to item 1.

According to the present invention, when the storage device is a RAID device, in order to give the highest priority to data guarantee, the shutdown processing is performed without performing the variable processing of the dirty data allowable amount between the occurrence of a power failure and the start of shutdown. To ensure the data.

  According to a third aspect of the present invention, there is provided a remaining battery capacity monitoring means for monitoring a remaining battery capacity of the storage device, and an allowable amount of dirty data in the cache memory of the storage device in accordance with the remaining battery capacity. Dirty data allowable amount control means for controlling, a means for performing a write-through operation when the allowable amount of dirty data is less than a threshold value, and a means for performing a write-back operation when the allowable amount of dirty data is greater than the threshold value And a dirty data processing device.

  The invention according to claim 4 of the present invention is characterized in that the storage device performs the shutdown process from the occurrence of a power failure to the start of the shutdown without performing the variable process of the dirty data allowable amount in the cache memory. The dirty data processing device according to claim 3.

  The invention according to claim 5 of the present invention is a dirty data processing program, comprising: a computer, a battery remaining capacity monitoring unit that monitors a remaining battery capacity of a storage apparatus; and the storage apparatus according to the remaining battery capacity. Dirty data allowable amount control means for controlling the allowable amount of dirty data in the cache memory possessed by, means for performing a write-through operation when the allowable amount of dirty data is less than a threshold, and the allowable amount of the dirty data is When it is larger than the threshold value, it is a dirty data processing program for functioning as a means for performing a write-back operation.

  According to the present invention, reliable data guarantee is ensured by the shutdown process after the occurrence of a power failure, and further, after the power is restored until the battery is fully charged, or when the battery device is in failure or deteriorated during normal operation. It is possible to operate the system while minimizing the decrease in response performance of the / O processing.

  The write-back operation is not performed until the dirty data becomes 0, but the threshold value is set for the remaining capacity of the battery, and the write-back operation and the write-through operation are selectively used according to the dirty data capacity. Compared with the case where the write-back operation is performed until the I / O processing response becomes unstable.

  FIG. 1 is an example of an apparatus according to the present invention. The RAID device 1 is connected to a host computer 2. The RAID device 1 includes a RAID controller 1-1, a backup disk 1-2, a disk 1-3, a cache memory 1-4, dirty data 1-5, a dirty data allowable area 1-6 during normal battery operation, and a battery 1-7. Have

  The RAID controller 1-1 is a controller that controls the exchange of information within the RAID device 1 and between the host computer 2 and the RAID device 1. The backup disk 1-2 is a backup disk device related to the RAID device 1, and dirty data 1-5 is saved when a power failure occurs. The disk 1-3 is a main disk device stored in the RAID device 1. The cache memory 1-4 is a cache memory of a disk device in the RAID device 1, and is composed of, for example, a high-speed semiconductor memory. The dirty data 1-5 is stored in the cache memory 1-4, and is overwritten on the corresponding information area in the disk 1-3 sequentially at a predetermined time. The normal dirty data allowable area 1-6 in which the remaining capacity of the battery is full is a dirty data write data area that is permitted to be used when the battery 1-7 of the RAID device 1 is fully charged. When the battery 1-7 is not fully charged, that is, when the battery 1-7 fails or is repaired, as shown in the dirty data 1-5, the dirty data allowable area 1-6 during normal battery operation It is operated with less capacity.

  FIG. 2 is an explanatory diagram of the dirty data capacity variable processing. Two types of MAX Dirty Pages, that is, Default MAX Dirty Page 2-1 and Current MAX Dirty Page 2-2 are stored inside the RAID controller 1-1 in FIG. Default MAX Dirty Page 2-1 is a configuration MAX Dirty Page, and the number of Dirty Pages that can be held when there is no abnormality in the remaining battery capacity is set. It is determined when the device is turned on, does not change even when the remaining battery capacity is abnormal, and is used to determine the value of Current MAX Dirty Page2-2. Current MAX Dirty Page 2-2 is the operational MAX Dirty Page, which is the same value as Default MAX Dirty Page 2-1 during normal operation, but decreases when battery remaining capacity is abnormal. used.

  The current Dirty Page number (hereinafter referred to as Dirty Count) is controlled by, for example, an internal counter and controlled so as not to exceed Current MAX Dirty Page 2-2. FIG. 2 shows Dirty Count 2-3 during normal operation with no abnormality in the remaining battery capacity.

Here, when the remaining battery capacity varies, the RAID controller shown in FIG. 1 is notified of what percentage of dirty pages can be backed up. The cache control unit changes the Current MAX dirty page as follows from the notified dirty page percentage.
Current MAX Dirty Page = Default MAX Dirty Page × Dirty Page percentage.
Here, if the DirtyPage percentage is low and the Current MAX Dirty Page is set low, the dirty data will be depleted and it will be difficult to guarantee operation. .

Processing is divided as follows according to the current MAX Dirty Page after the change.
If there is no change in Current MAX Dirty Page, no particular processing is performed.
Then, processing when Current MAX Dirty Page decreases will be described based on a processing example when Current MAX Dirty Page decreases in FIG.
First, for example, if the amount of dirty data that can be backed up is greatly reduced at one time due to a battery failure, for example, if one of the four batteries fails and the capacity decreases by 25% at a time, The cache memory data is written back to the target disk until the dirty data capacity that can be backed up with the remaining battery capacity is reached. Here, if the write-back operation is performed, I / O processing cannot be accepted at all until the data in the cache memory is written back to the target disk. Therefore, in the present invention, I / O processing can be accepted by combining write-back operation and write-through operation. Here, the write-through operation that shifts when the Dirty Page percentage falls below a certain threshold is "battery error write-through operation", and the Dirty Count exceeds Current MAX Dirty Page due to battery failure. The write-through operation will be referred to as “dirty exhaustion factor write-through operation”.

  In FIG. 3, when Current MAX Dirty Page3-4 is equal to or less than the threshold value due to the notification of the remaining battery capacity change, the process shifts to the write-through operation for the cause of battery abnormality. However, if the write-through operation for the cause of Dirty depletion has been performed until then, the write-through operation for the cause of Dirty depletion is once canceled and the operation shifts to the write-through operation for the cause of battery abnormality again. In addition, when the write-through operation for the battery abnormality factor was originally performed, the shift control is not performed.

  If Current MAX Dirty Page 3-4 is equal to or greater than the threshold even when the notification is received, and Dirty Count 3-3 is greater than Current MAX Dirty Page 3-4, the process shifts to write-through operation as a cause of Dirty depletion.

  Here, the process is completed when the current state is other than the write-through operation of the Dirty depletion factor, but when the current state is the write-through operation of the Dirty depletion factor, the processing to the target disk is performed until the Over Dirty Count becomes 0. Write-back is accelerated, and when the Over Dirty Count reaches 0, the write-through operation for the cause of Dirty exhaustion is canceled and the operation shifts to the write-back operation.

  A case where the Current MAX Dirty Page increases will be described based on a processing example when the Current MAX Dirty Page increases in FIG.

  When the write-through operation for the cause of the battery abnormality is performed at the time of receiving the change notification of the remaining battery capacity, the operation shifts to the write-back operation. However, if Current MAX Dirty Page4-4 does not exceed the threshold value, the transition control is not performed.

  If the write-through operation for the Dirty depletion factor has been performed at the time of receiving the notification, the write-through operation for the Dirty depletion factor is once canceled and the operation shifts to the write-back operation.

  If the Dirty Count 4-3 is larger than the Current MAX Dirty Page 4-4 at the time of receiving the notification, the process shifts to the write-through operation of the Dirty depletion factor.

  Here, the process is completed when the current state is other than the write-through operation of the Dirty depletion factor, but when the current state is the write-through operation of the Dirty depletion factor, the processing to the target disk is performed until 0ver Dirty Count becomes 0. Write-back is accelerated, and when the Over Dirty Count reaches 0, the write-through operation for the cause of Dirty exhaustion is canceled and the operation shifts to the write-back operation.

  As described above, in the present invention, since a predetermined threshold is set for the remaining battery capacity and the operation is shifted to the write-back operation and the write-through operation according to the dirty data amount, it is constant regardless of the remaining battery capacity. I / O response can be guaranteed. In addition, even if a battery abnormality occurs, a certain I / O response can be guaranteed, and in the event of a power failure, the data is not lost if a battery failure occurs. Data guarantee can also be performed.

An embodiment of the device according to the invention Illustration of dirty data capacity variable processing Example of processing when Current MAX Dirty Page decreases Example of processing when Current MAX Dirty Page increases

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 RAID apparatus 1-1 RAID controller 1-2 Backup disk 1-3 Disk 1-4 Cache memory 1-5 Dirty data 1-6 Dirty data permissible area at the time of normal battery 1-7 Battery 2 Host computer

Claims (5)

Monitoring the remaining battery capacity of the storage device;
Controlling the allowable amount of dirty data in the cache memory of the storage device in accordance with the remaining battery capacity;
When the amount of dirty data is less than or equal to a threshold, performing a write-through operation;
A step of performing a write back operation when an allowable amount of the dirty data is larger than a threshold;
A dirty data processing method comprising:   2. The dirty data processing method according to claim 1, wherein the storage device performs a shutdown process without performing a variable process of a dirty data allowable amount in the cache memory from a power failure occurrence to a shutdown start. Battery remaining capacity monitoring means for monitoring the remaining battery capacity of the storage device;
Dirty data allowable amount control means for controlling the allowable amount of dirty data in the cache memory of the storage device according to the remaining battery capacity;
When the allowable amount of dirty data is less than or equal to a threshold, means for performing write-through operation;
Means for performing a write-back operation when an allowable amount of the dirty data is larger than a threshold value.   4. The dirty data processing apparatus according to claim 3, wherein the storage device performs a shutdown process without performing a variable process of an allowable amount of dirty data in the cache memory from the occurrence of a power failure to the start of the shutdown. Dirty data processing program comprising:
Battery remaining capacity monitoring means for monitoring the remaining battery capacity of the storage device;
Dirty data allowable amount control means for controlling the allowable amount of dirty data in the cache memory of the storage device according to the remaining battery capacity;
When the allowable amount of dirty data is less than or equal to a threshold, means for performing write-through operation;
When the allowable amount of dirty data is larger than a threshold, as means for performing a write-back operation,
Dirty data processing program to make it function.

Images