What is the main purpose of this?

[mark988]

这个主要是什么用途？

[shashank-shark]

Consider a scenario where you have to do fast read and writes to a file in the case of sub-millisecond latency. Reading from file and writing to file (more system calls) under these scenarios itself adds on the extra delay or penalty.

Now as a solution to this issue would be typical virtual memory that operating system would offer for a process (here the java process which would read and write to file) , where the files is virtually mapped (physical memory address to virtual memory address mapping) so that a process directly has access to the file for read and write operations. This would decrease the number of system calls considerably.

In a similar way if you want to do some sort of caching for fast KEY,VALUE pair data access you might need a HashMap for that instead of a file. So this is the library that offers it.

TRANSLATED:

考虑以下情况，在亚毫秒级延迟的情况下，您必须快速读写文件。在这些情况下，从文件读取和写入文件（更多系统调用）本身会增加额外的延迟或代价。

现在，解决此问题的方法是操作系统将为进程（此处为将读写文件的java进程）提供给进程的典型虚拟内存，其中虚拟地映射了文件（物理内存地址到虚拟内存地址的映射）因此，进程可以直接访问该文件以进行读写操作。这将大大减少系统调用的次数。

以类似的方式，如果您想对快速的KEY，VALUE对数据访问进行某种缓存，则可能需要一个HashMap来代替文件。这就是提供它的图书馆。

[ieugen]

I just found out about this library.
We are looking for a way to load into memory very large hash maps structures from files ~100MB-2-3GB in size.

Do you have some benchmarks for this library?