dbms.buffer.BufferAccess

最新推荐文章于 2024-09-19 17:20:04 发布

shootogawa

最新推荐文章于 2024-09-19 17:20:04 发布

阅读量1k

点赞数

CC 4.0 BY-SA版权

文章标签： integer hashmap null byte 算法 schema

本文链接：https://blog.youkuaiyun.com/shootogawa/article/details/6292965

本文介绍了一个缓存管理系统的设计实现，包括缓存池管理、时钟淘汰算法、最近最少使用（LRU）算法以及如何处理锁定的关键块。系统采用单例模式，并支持不同类型的特殊块如元数据块、模式块等。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

public class BufferAccess{

public static final int cacheSize= 16*1024*1024/DiskManagement.BLOCK_SIZE;

private static BufferAccess bufferAccessInstance= null;

private DiskManagement diskManagement;

          /**
        * 锁住的关键块,只在文件打开时放入,文件关闭时取出
        * 存放mata, schema, index block, index info block
         */
          private Set LockedBlockSet = new HashSet();

          /**
          * 块缓冲池,load factor 0.75,确保没有rehash发生
          * 方便起见，只存DataBlock
          */

private Map blockCache=new HashMap((int)(cacheSize/0.75));

           /**
         * 时钟算法的环中的当前元素，下一个元素才是淘汰候选对象
          */
          private DataBlock lastUsedBlock;
          //stop here , review the clock algo and LRU algo

private BufferAccess(){

this.diskManagement = DiskManagement.getInstance();

}

public static BufferAccess getInstance() {

if(bufferAccessInstance == null)

bufferAccessInstance = new BufferAccess();

return bufferAccessInstance;

}

             /**
              * 从buffer中得到一个数据块
             * @param fileID
              * @param blockID
              * @return
              */

public DataBlock getBlock (Integer fileID, int blockID){

IDEntry id =new IDEntry (fileID,blockID);

DataBlock block =(DataBlock) blockCache.get(id);

if(block !=null){

block.keepNextTurn=true;

return block;

}

byte[] blockByte=diskManagement.readFromDisk(fileID,blockID);

block =new DataBlock(blockByte,id);

ensureSpace();

addBlock(block);

blockCache.put(id,block);

return block;

}

void reloadDataBlock(DataBlock nullBlock){

IDEntry id=nullBlock.getIdEntry();

DataBlock block =(DataBlock) blockCache.get(id);

//内存中没有相同id的块

if(block ==null)

{

byte[] blockByte=diskManagement.readFromDisk(id.getFileID(),id.getBlockID());

ensureSpace();

nullBlock.setData(blockByte);

addBlock(nullBlock);

blockCache.put(id,nullBlock);

return;

}

//内存中有相同id的块，需要用到identical指针

nullBlock.setData(block.getData());

nullBlock.identical=block.identical;

block.identical = nullBock;

}

                     /**
                       * 得到一个锁住的块，该块不参与LRU算法的竞争
                      * 只可能是MetaBlock,SchemaBlock,IndexBlock
                       * @param fileID
                       * @param blockID
                       * @param blockType
                       * @return
                       */

public Block getLockedBlock(Integer fileID, int blockID, int blockType) {

IDEntry id = new IDEntry(fileID, blockID);

Block block = null;

byte[] blockByte = diskManagement.readFromDisk(fileID, blockID);
switch(blockType){

case Block.META_BLOCK:

block= new MetaBlock(blockByte,id);

break;

case Block.SCHEMA_BLOCK:
block = new SchemaBlock(blockByte,id);

break;

case Block.INDEX_BLOCK:

block = new IndexBlock( blockByte, id );

break;

case Block.INDEX_INFO_BLOCK:

block = new IndexInfoBlock( blockByte, id);

break;

                                 default:
                                 System.out.println(" illegal block type argument ");
                                }

LockBlockSet.add(block);s