1 / 13

Container Marking : Combining Data Placement, Garbage Collection and Wear Leveling for Flash

Container Marking : Combining Data Placement, Garbage Collection and Wear Leveling for Flash. MASCOTS '11. Xiao-Yu Hu , Robert Haas, and Eleftheriou Evangelos From IBM Reserarch , Zurich, Switzerland. 2011. 08. 23 ( Tue ) Kwangwoon univ . SystemSoftware Lab. HoSeok Seo.

bree
Télécharger la présentation

Container Marking : Combining Data Placement, Garbage Collection and Wear Leveling for Flash

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. Container Marking : Combining Data Placement, Garbage Collection and Wear Leveling for Flash MASCOTS '11 Xiao-Yu Hu, Robert Haas, and EleftheriouEvangelos From IBM Reserarch, Zurich, Switzerland 2011. 08. 23 (Tue) Kwangwoonuniv. SystemSoftware Lab. HoSeokSeo

  2. Introduction • Garbage Collection • Reclaim blocks that include invalid page to reuse it as free blocks • Wear Leveling • Use blocks evenly to extends overall SSD endurance life • But, Wear Leveling makes • Unnecessary invalid pages • Unnecessary Writes ( for relocation of valid pages ) • In this paper • Proposes Container Marking Scheme to solve these problems

  3. Background • Garbage Collection • Greedy • Select victims that have less valid pages • Cost-benefit(FIFO) • Consider two parameters • time after last updated block • Number of valid pages • Select victims that have the biggest value

  4. Motivation • Why are active and inactive pages distinguished?

  5. Scheme Principle • If a block has the same active level pages, it has high chance to have invalid pages a lot. • Marker • Blocks have a marker indicating block’s endurance life. • Pages have a marker indicating page’s active level • Data placement • One block has the same active level pages. • More active pages is located more younger blocks

  6. Container-marking based data placement • Block marker range : 1 ~ 2L • Low marker : older blocks • High marker : younger blocks • The free block pool is organized by a priority queue in term of the remaining life count of the blocks. • The priority queue can be logically divided into 2L segments. • Once a free block has been filled up with data, it is removed from the free block pool and goes into one of 2L lists of occupied blocks depending on its marker

  7. Container-marking based data placement

  8. Modified greedy selection • Victim selection policy for garbage collection. • Prevent relatively younger blocks from being locked by inactive data. • To select j-th block, V : the number of valid page in a block E : remaining life count

  9. Page marker estimation • Basic • At high utilization, pages on average have a higher chance to be relocated than to be updated. • In this case, the markers gradually decrease and converge to lower values.

  10. Page marker estimationwith probabilistic marker

  11. Implementation • Java-based Simulator • Map Policy : page-level mapping • Container-marking meta data (32bits) • Remaining life count (20bit) • The number of valid pages within the block (8bits) • Container marker (4bits)

  12. Result

  13. Result

More Related