#available column displays the amount of physical memory that is still available to the application. # I think only after understanding some basic After understanding the concept, the above output can help us understand the memory status of the system. buff/cache
Let me ask a question first: buffer and cache should be two types of memory, but why does the free command change What about putting them together? Answering this question requires some preparation. Let us first clarify the meaning of buffer and cache.
buffer
In the operating system, it refers to buffer cache, which is generally translated as "buffer" in Chinese. To understand buffers, two other concepts must be understood: "sector" and
"piece". A sector is the smallest addressing unit of a device, also called a "hard sector" or "device block". Block is the smallest addressing unit of the file system in the operating system, also called "file block" or "I/O
"block". Each block contains one or more sectors, but cannot be larger than a page, so a page can hold one or more blocks in memory. When a block is loaded into memory, it is stored in a buffer area. Each buffer corresponds to a block, which is equivalent to the representation of a disk block in memory (the picture below is from the Internet):
Note that the buffer cache only has the concept of blocks and not files. It just moves the blocks on the disk directly into the memory without caring about what format of files are stored in the blocks.
cache refers to page cache in the operating system, and Chinese is generally translated as "page cache". The page cache is a disk cache implemented by the kernel. It is mainly used to reduce the impact on the disk. I/O operations. Specifically, by caching the data in the disk into physical memory, access to the disk is changed into access to physical memory. The page cache caches memory pages. In the cache Pages come from reading and writing ordinary files, block device files (this refers to the buffer cache) and memory mapped files.
We can understand the caching of ordinary files by the page cache in this way: when the kernel To read a file (such as
/etc/hosts), it will first check whether the data of this file is already in the page cache. If so, give up accessing the disk and read directly from memory. This behavior is called a cache hit. If the data is not in the cache, it means a cache miss, and the kernel will schedule the block.
I/O operations read data from disk. The kernel then places the read data into the page cache. This cache targets files recognized by the file system (such as /etc/hosts).
The page cache's cache of block device files is the buffer cahce we introduced earlier. Because individual disk blocks are also stored in the page cache through the buffer (the buffer is ultimately hosted by the page cache).
We should be clear at this point: whether it is a buffer or a page cache, they are implemented in the same way. The buffer is just a conceptually special page cache.
Then why
Isn't the free command directly called cache instead of buff/cache? This is because buffer and page cache implementations are not inherently unified. in linux
Kernel 2.4
Only then can they be unified. There were two separate disk caches in earlier kernels: the page cache and the buffer cache. The former caches pages and the latter caches buffers. Once you know the story, the names of the columns in the output may no longer matter.
free 与 available
在 free 命令的输出中,有一个 free 列,同时还有一个 available 列。这二者到底有何区别?
free
是真正尚未被使用的物理内存数量。至于 available 就比较有意思了,它是从应用程序的角度看到的可用内存数量。Linux
内核为了提升磁盘操作的性能,会消耗一部分内存去缓存磁盘数据,就是我们介绍的 buffer 和 cache。所以对于内核来说,buffer 和
cache 都属于已经被使用的内存。当应用程序需要内存时,如果没有足够的 free 内存可以用,内核就会从 buffer 和 cache
中回收内存来满足应用程序的请求。所以从应用程序的角度来说,available = free + buffer + cache。请注意,这只是一个很理想的计算方式,实际中的数据往往有较大的误差。
交换空间(swap space)
swap
space 是磁盘上的一块区域,可以是一个分区,也可以是一个文件。所以具体的实现可以是 swap 分区也可以是 swap
文件。当系统物理内存吃紧时,Linux 会将内存中不常访问的数据保存到 swap 上,这样系统就有更多的物理内存为各个进程服务,而当系统需要访问
swap 上存储的内容时,再将 swap
上的数据加载到内存中,这就是常说的换出和换入。交换空间可以在一定程度上缓解内存不足的情况,但是它需要读写磁盘数据,所以性能不是很高。
现在的机器一般都不太缺内存,如果系统默认还是使用了
swap 是不是会拖累系统的性能?理论上是的,但实际上可能性并不是很大。并且内核提供了一个叫做 swappiness
的参数,用于配置需要将内存中不常用的数据移到 swap 中去的紧迫程度。这个参数的取值范围是 0~100,0 告诉内核尽可能的不要将内存数据移到
swap 中,也即只有在迫不得已的情况下才这么做,而 100 告诉内核只要有可能,尽量的将内存中不常访问的数据移到 swap 中。在
ubuntu 系统中,swappiness 的默认值是 60。如果我们觉着内存充足,可以在 /etc/sysctl.conf 文件中设置
swappiness:
如果系统的内存不足,则需要根据物理内存的大小来设置交换空间的大小。具体的策略网上有很丰富的资料,这里笔者不再赘述。
/proc/meminfo 文件
其实 free 命令中的信息都来自于 /proc/meminfo 文件。/proc/meminfo 文件包含了更多更原始的信息,只是看起来不太直观:
有兴趣的同学可以直接查看这个文件。
总结
free 命令是一个既简单又复杂的命令。简单是因为这个命令的参数少,输出结果清晰。说它复杂则是因为它背后是比较晦涩的操作系统中的概念,如果不清楚这些概念,即便看了 free 命令的输出也 get 不到多少有价值的信息。
相关推荐:《Linux视频教程》