CN101221465A - Data buffer implementation method for reducing hard disk power consumption - Google Patents

Abstract

The method for realizing the data buffer for reducing the power consumption of the hard disk is applicable to the computer storage system, and the hard disk data is cached by using the data buffer, and the average power consumption of the hard disk is reduced through the innovative design of the buffer structure in a multi-tasking environment. The buffer includes a block data buffer and a buffer management module. The block data buffer transmits data to the buffer management module, and the buffer management module transmits state control and data access commands to the hard disk through the hardware device driver module , and then the hard disk transmits the data to the block data buffer to realize the buffer for reducing the power consumption of the hard disk. On the one hand, according to the multi-task data access rate, a method to determine the buffer group size with optimal power consumption is proposed; on the other hand, a data block replacement algorithm considering task data access is proposed to further extend the idle time of hard disk access.

Description

Data buffer implementation method for reducing hard disk power consumption

technical field

The invention is applicable to a computer storage system, uses a data buffer to buffer hard disk data, and reduces the average power consumption of the hard disk through the innovative design of the buffer structure in a multi-task environment.

Background technique

In the computer system, the buffer (Buffer) is a very widely used technical means. A buffer refers to an area used to store data transferred between devices with asynchronous speeds or devices with different priorities. Therefore, traditional buffers are designed for performance.

In a computer system, the power consumption of a hard disk device is relatively high. In consideration of saving energy, generally speaking, a hard disk has multiple power consumption states. When the hard disk is idle, it can enter a low-power state to save energy. The hard disk will generate energy loss when entering and exiting the low power consumption state, so only entering the low power consumption state when the hard disk is idle for a long time can really save energy.

Data prefetching (Prefetch) can be performed on the hard disk by using the buffer zone, that is, a large amount of data is read from the hard disk to the buffer zone each time. Therefore, the hard disk will generate a long idle time between each read-ahead. Even considering the power consumption of the buffer itself, putting the hard disk into a low power consumption state during this long period of time can save considerable energy. In this way, buffer technology can be used to reduce the average power consumption of the hard disk.

The premise of using this technology to reduce the power consumption of the hard disk is to access the data sequentially. However, in a multi-tasking environment, the data access of the actual system cannot be performed in a purely sequential manner: ①Due to the influence of task scheduling, multi-task access is interleaved; ②For a single task, the task may affect the hard disk. The data segment performs random access: each access starts from a random position in the data segment, and continuously accesses the following segment of data. Since the existing buffer technology does not retain the pre-read content of each task, the existence of the above factors will cause the content of the buffer to be replaced frequently, resulting in a large number of hard disk accesses, thereby greatly shortening the idle time and reducing the power consumption of the hard disk. get lowered.

Contents of the invention

Technical problem: The purpose of this invention is to provide a data buffer design method for reducing hard disk power consumption. In a multi-tasking environment, by dividing the buffer into blocks, each task can retain the data content pre-read from the hard disk in different groups composed of buffer blocks, so that the pre-read data read from the hard disk will not be due to tasks. Frequent switching out of the hard disk has the effect of prolonging the idle time of the hard disk and reducing energy consumption.

Technical solution: In order to reduce the power consumption of the hard disk through the buffer in a multi-task environment, the pre-read data of each task should be reused to avoid frequent access to the hard disk. The invention divides the buffer into blocks, and groups the buffer blocks during operation, so that each task retains its own pre-reading content in different groups. According to the status of data access, the status of the buffer block is marked, and the replacement algorithm selects the appropriate block to replace according to this mark.

The present invention is used to reduce the hard disk power consumption buffer realization method specifically: the buffer comprises the data buffer of block, buffer management module, the data buffer of block transmits data to the buffer management module, the buffer management module The state control and data access commands are transmitted to the hard disk through the hardware device driver module, and the hard disk then transmits the data to the block data buffer to realize the buffer for reducing the power consumption of the hard disk.

The data buffer of described block is: open up the zone of fixed size in memory as data buffer, size is D byte; Data buffer is divided into fixed block number N, and the size of every block is equal, is S word section, D=S×N; initially there is no data in each block, and it is connected through a linked list (Tsinghua University Press "C Programming", written by Tan Haoqiang) L1, and each structure of the linked list L1 has two members, the first One stores buffer block numbers 1 to N, and the second stores the first address of the next buffer block in memory.

The buffer management module is: there are M tasks, and the task obtains the required data by calling the interface function FD provided by the hard disk drive; the buffer management module provides a function with the same name, the same parameters and the same return value as the hard disk drive interface function FD FM, in the task code, replace the hard disk interface function FD with the function FM provided by the buffer management module; the function FM is implemented as follows: starting from the first address of the task’s request for hard disk data access, the length of the pre-read from the hard disk is S×N/ M bytes of data, and select the buffer block used to store the data from all buffer blocks through the replacement algorithm R; connect these buffer blocks through the linked list L2, and the connected buffer blocks are called groups, which will soon belong to The buffer blocks of different tasks are connected through the linked list L2; when the task calls the function FM to read data from the hard disk, if the data already exists in the buffer block, the buffer management module searches for the data in each buffer block through the linked list L2 And deliver the task of initiating the request; when each pre-reading is completed, the buffer management module sends a command to the hard disk to enter the low power consumption mode.

The replacement algorithm R is:

Determine different states for buffer blocks 1 to N, these states include: buffer block idle (BLK_IDLE), buffer block in use (BLK_INUSE), buffer block used (BLK_USED), buffer block unused (BLK_UNUSE) ; The buffer block is free, indicating that no data has been read in the buffer block; the buffer block is in use, indicating that the data in the buffer block is being read; the buffer block is used, indicating that the last byte of data in the buffer block has been is read; the buffer block is not used, which means that the buffer block already exists in the group but has not been read; the replacement algorithm sequentially queries the status of the buffer block in each group through the linked list L2, if there is a buffer whose state is that the buffer block is free block, then read the data into the block; if there is no buffer block whose status is buffer block free, then query whether there is a buffer block whose status is buffer block used, and if so, read the data into the buffer block; if the buffer block whose status is buffer block used does not exist, then query whether there is a buffer block whose status is buffer block unused, and if it exists, read the data into the buffer block; if the buffer block The block is free, the buffer block is used, and the buffer block is not used. The buffer block in the three states does not exist, and the data is read into the buffer block in use by the buffer block. If multiple buffer blocks with the same status are found during the query process, the first one found in the query is selected.

Beneficial effects: the invention logically groups the buffers so that each task accessing the hard disk can have its own data buffer. The present invention does not need to add any hardware circuit, and the overhead of computing and storage resources is extremely small. In a multi-tasking environment, the data access request to the hard disk is greatly reduced, and the idle time of the hard disk is extended, so that the hard disk can be used for a longer period of time. Enter a low-power state to save energy. Experimental results show that in the application environment of multimedia playback, the power consumption of hard disk data access has been greatly reduced.

Description of drawings

FIG. 1 is a basic structural diagram of the design architecture of the low power consumption scheme of the hard disk based on buffer dynamic grouping in the present invention. The solid line arrows in the figure indicate data requests and control commands, and the dotted line arrows indicate data. The buffer management module receives the data access request generated by the task to the hard disk, pre-reads the data from the hard disk into the buffer through the hard disk device driver, and submits the data required by the task to the task that initiated the request; when each pre-read is completed, the buffer The area management module updates the group data structure of the data buffer and puts the hard disk in a low power consumption state through the hard disk device driver.

FIG. 2 is a structure of a linked list L1 for linking buffer blocks without data. Each structure of the linked list L1 has two members, the first one stores buffer block numbers 1-N, and the second one stores the first address of the next buffer block in memory. The linked list searches for the next buffer block through the first address of the next buffer.

Figure 3 illustrates the linked list L2 used to link buffer groups. The solid line frame represents the buffer block (Block), and the number in it is the block number; the dotted line frame represents the buffer group (Group), in which each block stores hard disk data with continuous addresses, and the data in each group belongs to the same task. The buffer blocks in each group are connected through the same linked list structure as the linked list L1, and each group is connected through the linked list L2. Each structure of the linked list L2 has two members, the first one stores the group number, and the second one stores the first address of the first buffer block in the next group.

Figure 4 is a flowchart of the implementation of the buffer management module. The buffer management module is the core module of the present invention, and Fig. 4 has described the detailed work flow of this module.

Detailed ways

The buffer includes a block data buffer and a buffer management module. The block data buffer transmits data to the buffer management module, and the buffer management module transmits state control and data access commands to the hard disk through the hardware device driver module , and then the hard disk transmits the data to the block data buffer to realize the buffer for reducing the power consumption of the hard disk.

The buffer is organized according to the structure shown in Figure 3. The block size of the buffer zone can be determined according to the actual situation of the system, and the total block number M is the sum of each pre-read block number of each task (the pre-read block number of each task can be determined according to the task situation). As shown in FIG. 3 , the solid arrow indicates that the group structure of the buffer is implemented through a linked list, which is a secondary linked list. The first address of the first block in each buffer group is connected through a first-level linked list, and the first addresses of each block in each group are connected through a second-level linked list.

The buffer management module is implemented between the application layer and the driver layer. This module provides the interface function of hard disk data access to the task to shield the implementation of the lower layer. Through this interface, the task informs the buffer management module of the start address and length of the data to be accessed, and each time the task calls the data access interface, a data request is generated. At this time, the buffer management module operates according to the process shown in FIG. 4 . This module completes data operation and state control by calling the interface function of the hard disk device driver. The module searches for data blocks by traversing the secondary linked list shown in Figure 3.

The data buffer of described block is: open up the zone of fixed size in memory as data buffer, size is D byte; Data buffer is divided into fixed block number N, and the size of every block is equal, is S word section, D=S×N; there is no data in each block at the beginning, and they are connected through the linked list L1. Each structure of the linked list L1 has two members, the first one stores the buffer block numbers 1~N, and the second one stores The first address of the next buffer block in memory.

Described buffer management module is: be provided with M task, task obtains required data by calling the interface function FD that hard disk drive provides; The function FM of the value, in the task code, replace the hard disk interface function FD with the function FM provided by the buffer management module; the function FM is implemented as follows: starting from the first address of the task’s request for hard disk data access, the length of the pre-read from the hard disk is S ×N/M bytes of data, and select the buffer block used to store the data from all buffer blocks through the replacement algorithm R; connect these buffer blocks through the linked list L2, and the connected buffer blocks are called grouping , that is, the buffer blocks belonging to different tasks are connected through the linked list L2; when the task calls the function FM to read data from the hard disk, if the data already exists in the buffer block, the buffer management module uses the linked list L2 in each buffer block The task of searching for data and sending it to the request; when each read-ahead is completed, the buffer management module sends a command to the hard disk to enter a low power consumption mode.

In the replacement algorithm, Block is divided into different states, these states include: BLK_IDLE, BLK_INUSE, BLK_USED, BLK_UNUSE. BLK_IDLE indicates that no data has been read in the Block; BLK_INUSE indicates that the data in the Block is being used, and BLK_USED indicates that the last byte of data in the Block has been accessed (indicating that the Block has been accessed), and the two are closely related: any access to the Block All tasks will set the Block state to BLK_INUSE and add 1 to the SHARE_NUM value (SHARE_NUM indicates the number of tasks accessing the Block, the initial value is 0), when the task finishes accessing the last data in the Block, reduce the SHARE_NUM value by 1 and Check the value, if it is not 0 (indicating that there are still tasks accessing the block), keep the BLK_INUSE state unchanged, and if it is 0 (indicating that all tasks that have accessed the block before have been accessed), set the block state to BLK_USED; BLK_UNUSE indicates that the block Already exists in the group but has not been accessed by the task. The replacement algorithm sequentially queries the attributes of Blocks in each Group through the secondary linked list structure in Figure 3, and selects Blocks for replacement according to the priorities of BLK_IDLE, BLK_USED, BLK_UNUSE, and BLK_INUSE (from high to low). If multiple Blocks are in the same state, then Select the first one found. From the analysis of the meaning of the state, it can be seen that the probability that the Block with the state of BLK_USED, BLK_UNUSE, and BLK_INUSE will be used in the future will increase in turn, so this replacement algorithm can bring fewer Block updates.

Claims (4)

1. buffer zone implementation method that is used to reduce hard disk power consumption, it is characterized in that this buffer zone comprises data buffer, the buffer management module of piecemeal, the data buffer of piecemeal sends data to the buffer management module, the buffer management module sends State Control and data access command to hard disk by the hardware device drivers module, hard disk sends data to the data buffer of piecemeal again, realizes being used to reduce the buffer zone of hard disk power consumption.

2. the buffer zone implementation method that is used to reduce hard disk power consumption according to claim 1, the data buffer that it is characterized by described piecemeal is: the zone of opening up fixed size in storer is as the data buffer, and size is the D byte; Data buffering is divided into fixing piece count N, every equal and opposite in direction is S byte, D=S * N; No datat in each piece links to each other by chained list L1 when initial, has two members in each structure of chained list L1, and first deposits buffer blocks numbering 1～N, deposits the first address of next buffer blocks in internal memory for second.

3. the buffer zone implementation method that is used to reduce hard disk power consumption according to claim 1 is characterized by described buffer management module and is: be provided with M task, task is obtained desired data by the interface function FD that calls hard drive and provide; The buffer management module provides with hard disk drive interface function F D to be had of the same name, same parameter and with the function F M of rreturn value, in task code hard-disk interface function F D is replaced with the function F M that the buffer management module provides; Function F M is achieved as follows: from the request first address of task to the hard disc data visit, reading length in advance from hard disk is S * N/M byte data, and selects to be used to deposit the buffer blocks of these data from all buffer blocks by replacement algorithm R; These buffer blocks are connected by chained list L2, and the buffer blocks after the connection is called grouping, and the buffer blocks that is about to belong to different task connects by chained list L2 separately; When task call function F M carried out data read to hard disk, if data have been present in the buffer blocks, the buffer management module was searched data and is sent to the initiation tasks requested by chained list L2 in each buffer blocks; When run through in advance at every turn finish after the buffer management module give an order to hard disk and make it enter low-power consumption mode.

4. the buffer zone implementation method that is used to reduce hard disk power consumption according to claim 1 is characterized by described replacement algorithm R and is:

For buffer blocks 1～N determines different states, these states comprise: the buffer blocks free time, buffer blocks is in usefulness, and buffer blocks is usefulness, and buffer blocks is not used; Do not read in any data as yet in the idle expression of the buffer blocks buffer blocks; Buffer blocks represents that in usefulness the data in the buffer blocks are read; Buffer blocks is with representing that the data of last byte are read in the buffer blocks; Buffer blocks is with representing that buffer blocks has been present in the grouping but is not read as yet; Replace the state of algorithm,, then data are read in this piece if existence is the buffer blocks of buffer blocks free time by buffer blocks in each grouping of chained list L2 sequential query; If existence is not the buffer blocks of buffer blocks free time, inquire about then whether existence is the buffer blocks buffer blocks of usefulness, then data are read in this buffer blocks as existing; If state be buffer blocks the buffer blocks of usefulness do not exist yet, inquire about then whether existence is the not buffer blocks of usefulness of buffer blocks, then data are read in this buffer blocks as existing; If the buffer blocks free time, buffer blocks is usefulness, and buffer blocks does not exist with the buffer blocks of three kinds of states, then data is read in the buffer blocks of buffer blocks in usefulness.If in query script, find to have the same a plurality of buffer blocks of state, then select first to inquire.

Images