1 / 10

Public Caching

Public Caching. Michael Albrecht Rory Carmichael. Motivation. Distributed filesystems are useful in a variety of circumstances Active Storage Co-locality of some files important Two big problems encountered Network Latency Resource Contention. Potential Solutions. Whole-File Caching

nziemer
Télécharger la présentation

Public Caching

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. Public Caching Michael Albrecht Rory Carmichael

  2. Motivation • Distributed filesystems are useful in a variety of circumstances • Active Storage • Co-locality of some files important • Two big problems encountered • Network Latency • Resource Contention

  3. Potential Solutions • Whole-File Caching • Copies only help one person/machine • Replication • Wasteful • Unnecessary copies of unused files • Not enough copies of heavily used files • Maintaining consistency causes performance problems • Caching + Replication • Combines Problems from both

  4. Solution – Public Caches • Use Whole File Caching as normal • Use AFS-like consistency semantics • Publish location of your copy in the file’s “inode” • Provides usage-based replication • Reduction in unnecessary copies • Semantics • Copy On Read • Clobber On Close

  5. Implementation – Multi-FS • Parrot • Type II VM • Transparent Remote Filesystem access • Home of Multi-FS • Chirp • Distributed Filesystem used by Multi-FS • Log-style Metadata • Cache Cleaner • Removes out-of-date versions • Reduces log to minimum necessary

  6. application open A@ex/foothen read into memory parrot open A/ex/root/foo open B/ex/data/abc return “B, /ex/data/abc” B(data only) A (metadata& data) return “Hello World 2” • B/ex/data/ • abc: “Hello World 2” • A/ex/root/ • foo: B, /ex/data/abc • bar: A, /ex/data/xyz • A/ex/data/ • xyz: “Hello World” Multi-FS

  7. Multi-FS w/Public Caches client application • client/multicache/ • mno: “Hello World 2” return “Hello World 2” open A@ex/foothen read into memory parrot open A/ex/root/foo open B/ex/data/abc return “B, /data/ex/abc” B(data only) A (metadata& data) • A/ex/root/ • foo: B, /data/ex/abc • client, /multicache/mno • bar: A, /data/ex/xyz • A/ex/data/ • xyz: “Hello World” • B/ex/data/ • abc: “Hello World 2”

  8. Evaluation • Overhead • Not substantially different • Clustered File Requests • Impact of File Write

  9. Conclusions • Public Caching Works • Public Caching provides performance benefits • Public Caching overhead is not excessive

  10. Questions?

More Related