Geek Answers Handbook

Is it true that database management systems usually bypass file systems ..?

As I understand it, typical database management systems bypass the file system correctly? I understand that they manage their own disk space, and they write actual data and indexing systems, such as tree B, directly to disk blocks, bypassing any intermediate help from the file system.

This assumes that root will give the user access to the database for direct reading and writing from disk blocks. On Linux, this is still simpler because the disk can be thought of as a file.

Any reference to real case studies is welcome.

+5
source share
3 answers

Most rely on the core file system for WAL , etc.: basically, they transfer it to the OS.

Support for some DBMSs (Oracle, MySQL) is "raw", but this is not typical. Too much trouble ( see this postgres chat ) because you still need WAL, etc. On your raw section.

+5
source

. , Windows Linux ( MySQL). , , (, / ) . . - - Varnish , , , .

+2

, mysql , , , .

Moreover, it may also differ from one storage mechanism to another. Mongodb uses memory mapped files for an IO disk. We look forward to further discussion here.

+1
source

More articles:


All Articles