• <ins id="pjuwb"></ins>
    <blockquote id="pjuwb"><pre id="pjuwb"></pre></blockquote>
    <noscript id="pjuwb"></noscript>
          <sup id="pjuwb"><pre id="pjuwb"></pre></sup>
            <dd id="pjuwb"></dd>
            <abbr id="pjuwb"></abbr>

            那誰的技術(shù)博客

            感興趣領(lǐng)域:高性能服務(wù)器編程,存儲(chǔ),算法,Linux內(nèi)核
            隨筆 - 210, 文章 - 0, 評(píng)論 - 1183, 引用 - 0
            數(shù)據(jù)加載中……

            The design and implementation of CommonCache Library(base on version 0.6)


            1. Overview

            a) the functions of a cache
            cache the most frequently visited data in the memory, so:
            1) if the the data get losed is not a big matter,it means that you should not store data
            in cache only, offen, cache is between the DB server and application,hence reduce visiting DB times
             and speed up application.
            2) since the memory is not quite big, so the eliminating algorithm such LRU must to be used.
            3) must support key-value-like API such as find/erase/set and so forth.

            b) ccache feature (version 0.6):
            1) using mmap to cache data in file, so it can be used in multi-thead and multi-process application.
            2) support unfix size key node
            3) support hash-rbtree and hash-list structure
            4) use LRU algorithm to eliminate nodes when it is running out of its memory size
            5) fast, 100W/read operation on fix-size key node in no more than 0.5 second

            c) what is the difference between ccache and memcached?
            ccache is a static library, memcached is a cache sever, so users who use ccache must
            design the protocols between clients and server, and write the server application themself.

            2. implementation
            a) the cache file structure


            b) the hashitem array detail   
            the hashitem array is used to find the node base on the node key.Every nodes in a
            same hashitem has a same hash num.All the nodes in the same hashitem are strutured
            in list or rbtree:


            when compiling the ccache, define the macro CCACHE_USE_RBTREE to specily using rbtree,
            else macro CCACHE_USE_LIST using list. The structrue ccache_node_t has following fields
            about this:

            typedef struct ccache_node_t
            {
                    
            /* 
                      
                     
            */

            #ifdef CCACHE_USE_LIST
                
            struct ccache_node_t *next, *prev;
            #elif defined CCACHE_USE_RBTREE    
                ccache_color_t color;
                
            struct ccache_node_t *parent, *left, *right;
            #endif

                    
            /* 
                      
                     
            */

            }
            ccache_node_t;

            the hashitem array size is defined in the configure file when open the cache.

            c) the freearea array detail   
            ccache use slab-like algorithm to allocate node memory.In the configure file, there is an
            item called "alignsize", this value specify the align size between different freearea,
            assume the alignsize is 8 bytes:

            all the nodes in the same freearea has the same size, when allocating a node(ccache_memory.c/ccache_allocate):
            1) allign the node size with the alignsize value, then find the fit
            freearea(ccache_memory.c/ccache_get_freeareaid) and return the freeareaid
            2) if the data zone has enough memory for the node, change the cache->start_free and cache->freesize
            and return the node pointer
            3) else, erase the  cache->freearea[freeareaid]->lrulast node, use this node memory
            for this allocation.
            4) every time a node allocated, place it in the freearea's tail(freearea->lrulast = node),
            when the node has been visited, it move on one step in the freearea list.So, the more a node has been
            visited, the closer it moves to the freearea head(freearea->lrufirst).Hence, the freearea lrulast
            node is the less frequently visited node in the freearea(although not so accurate:).

            the ccache_node_t structure has following fields about LRU:

            typedef struct ccache_node_t
            {
                
            /*
                   .
                
            */
                   
                
            struct ccache_node_t *lrunext, *lruprev;

                
            /*
                   .
                
            */
                   
            }
            ccache_node_t;

             

             

            posted on 2009-10-22 21:11 那誰 閱讀(4118) 評(píng)論(1)  編輯 收藏 引用 所屬分類: ccache

            評(píng)論

            # re: The design and implementation of CommonCache Library(base on version 0.6)  回復(fù)  更多評(píng)論   

            開源,大力支持啊!
            中國缺少這樣的東西
            2010-03-10 19:37 | jesse
            久久AV高潮AV无码AV| 久久精品水蜜桃av综合天堂| 久久人人爽人人爽AV片| 久久精品一区二区三区AV| 久久人人爽爽爽人久久久| 国产真实乱对白精彩久久| 久久久国产精华液| 久久99国产精品一区二区| 久久无码精品一区二区三区| 久久久无码一区二区三区| 久久久久女教师免费一区| 久久人爽人人爽人人片AV| 久久综合日本熟妇| 久久亚洲国产精品一区二区| 久久天天躁夜夜躁狠狠| 久久久久久亚洲精品无码| 国产精品va久久久久久久| 亚洲人成伊人成综合网久久久| 66精品综合久久久久久久| 国产69精品久久久久APP下载| 亚洲国产成人久久精品影视| 欧美激情一区二区久久久| 久久久久亚洲精品男人的天堂| 精品国产乱码久久久久久1区2区| 偷偷做久久久久网站| 久久久久久久亚洲精品| 国产精品成人精品久久久| 久久精品蜜芽亚洲国产AV| 中文字幕热久久久久久久| 午夜精品久久久久久久无码| 精品国产综合区久久久久久 | 无码精品久久久天天影视| 色8激情欧美成人久久综合电| 国产日韩久久免费影院| 国产精品内射久久久久欢欢| 久久精品男人影院| 国产激情久久久久影院| 一本一道久久精品综合| 国产69精品久久久久9999| 91精品国产91久久久久久| 国产精品99久久不卡|