Arranging Pages of a File on Disk
- 'Next' page concept:
Pages on same track, followed by pages on same cylinder, followed by pages on adjacent cylinder. - Pages in a file should be arranged sequentailly by 'next',to minimize seek and rotational delay.
- For sequential scan, pre-fetching several pages at a time is a big win!
- In practice, data pages not always sequentially stored on disk, why?
-> Database grows and shrinks
한 디스크의 트랙을 다 쓰고나면, 플래터의 반대면 같은 트랙을 쓰고, 그리고 다쓰면 다른 플래터의 윗면..(최대한 같은 실린더를 쓴다) 그러고도 다쓰면 맨처음 다쓴 디스크의 바로 안쪽 트랙을 사용한다. 이렇게 함으로써 seek time을 줄일 수 있다.
Different Layers
Files
- For user, a file is searched or updated as a collection of logical records.
- Example: find student records in CS major
- For DBMS, a file is stored, read and written as a collection of disk pages.
- Example: read data identified byt page ids 1, 4, 10 (where student records in CS major are stored)
- Memory/disk managers must translate user's search or update into disk page access, done by indexing.
Disk Management
- Mange the space on disk
- Disk pages are uniquely identified by page id(b,t,c,d), "block of track t of cylinder c of disk d"\
- Higher level(i.e., memory manger) calls upon this layer to
- Allocate / de-allocate a page for a file
- Read/write a page
- Keep track of data pages and free pages
- Higher levels don't need to know how this is done.
Memory (RAM) Management
When a page is requested ...
- If requested page is not in buffer pool(memory - by Table lookup), i.e.,page fault:
- Choose a frame for replacement
- If frame is dirty(i.e., modified but not written to disk), write it to disk
- Read requested page into chosed frame and update Table
- Pin the page and return its address
More on Buffer Management
- Requestor of page must unpin it and indicate whether page has been modified:
- dirty bit is used for this
- Page in pool may be requested many times (by many requestors)
- a pin count is used. A page is a candidate for replacement iff pin count = 0
(pin은 사용자가 사용중에 있다는 것을 의미한다)
(dirty는 modified 됐다는 것을 의미한다)
