1 / 18

CS422 Principles of Database Systems Buffer Management

CS422 Principles of Database Systems Buffer Management. Chengyu Sun California State University, Los Angeles. Memory Hierarchy. CPU. Cache. Memory. Disk. Buffers in a Computer. Disk cache Memory buffer L1, L2, and L3 caches. Why OS Memory Buffer Is Not Enough.

christopherramirez
Download Presentation

CS422 Principles of Database Systems Buffer Management

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. CS422 Principles of Database SystemsBuffer Management Chengyu Sun California State University, Los Angeles

  2. Memory Hierarchy CPU Cache Memory Disk

  3. Buffers in a Computer • Disk cache • Memory buffer • L1, L2, and L3 caches

  4. Why OS Memory Buffer Is Not Enough • DBMS knows its data better • Database buffer management must be coordinated with failure recovery mechanisms

  5. Data Access Without Buffer Management • Load a block into a page • Access data in the page • Write the page back to disk if the data is changed • Release the page So why do we need buffer management??

  6. Buffer Management – Buffer Manager • A buffer manager manages a fixed set of pages, called a buffer pool • Each page in the buffer pool is called a buffer page

  7. Buffer Management – Client Code • Processes access disk through buffer manager • These processes are referred to as client code (or just client)

  8. Buffer Management – Pin • Buffer pin(Block) • Load a block into a buffer page • Mark the buffer page as pinned • A pinned buffer page is being used by some client code • A unpinned buffer page is available for reuse

  9. Four Possible Cases for Pin • The block to be pinned is already buffered in memory • The buffer is pinned • The buffer is not pinned • The block to be pinned is not buffered in memory • There are unpinned buffers available • All buffers are pinned

  10. Buffer Management – Read/Write Data • If the data in a page is changed, the page is called a dirty page

  11. Buffer Management – Unpin • unpin(Buffer) • Indicates the page is no longer used by the client

  12. Buffer Management – Flush • Write the dirty page(s) to disk • When to flush • Before the page is pinned to a different block • At the request of the failure recovery mechanism

  13. Example: Buffer Replacement • Size of buffer pool: 4 • What does the buffer pool looks like after the following requests: pin(0), pin(1), pin(2), pin(3), unpin(3), unpin(1), unpin(2),pin(5) Buffer Pool

  14. Buffer Replacement Policies • Naïve • Sequentially scan the buffer pool and replace the first unpinned page • Clock • FIFO (First In First Out) • LRU (Least Recently Used)

  15. Problem of the Naïve Policy • pin(1), unpin(1), pin(2), unpin(2), pin(1), unpin(1), pin(2), unpin(2)…

  16. Clock Policy • Sequentially scan the buffer pool and choose the first unpinned page • Start the next scan at the page after the previous replacement

  17. Implementing FIFO and LRU • FIFO • For each buffer page, keeps the time when the block is read in • LRU • For each buffer page, keeps the time when the page is unpinned

  18. Readings • Chapter 13.4 and 13.5 of the textbook

More Related