File metadata synchronizing access managing file

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Data shipping mode is primarily used by the MPI/IO library –  MPI/IO does not require POSIX seman,cs –  MPI/IO library provides a natural mechanism to define and assign blocks to nodes –  Avoids the overhead of flushing dirty data to disk when revoking a token 1 2/4/13 CS480 Principles of Data Management Spring 2013 CS480 Principles of Data Management Spring 2013 2. File Metadata: Synchronizing Access Managing File Metadata GPFS •  GPFS uses inodes and indirect blocks to store file aFributes and data block addresses •  Mul,ple nodes wri,ng to the same file result in concurrent updates to the inode and indirect blocks of the file –  To change file size and modifica,on ,me (m,me) and to store the addresses of newly allocated data blocks •  Update via exclusive write locks on the inode will result in a lock conflict on every write opera,on CS480 Principles of Data Management Spring 2013 Shared write lock (1/2) Spring 2013 Shared write lock (2/2) •  On inode •  Allows concurrent writers on mul,ple nodes •  This shared write lock only conflicts with opera,ons that require the exact file size and/or m,me –  a stat() system call or a read opera,on that aFempts to read past end ­of ­file •  One of the nodes accessing the file is designated as the metanode for the file –  Only the metanode reads or writes the inode from or to disk –  Centralized approach CS480 Principles of Data Management CS480 Principles of Data Management Spring 2013 Updates to indirect blocks •  Each writer maintains a locally cached copy of the inode and forwards its inode updates to the metanode: –  Periodically or –  Every ,me the shared write token is revoked •  Metanode merges inode updates –  Retaining the largest file size and latest m,me values –  Opera,ons that update file size or m,me requires an exclusive inode lock CS480 Principles of Data Management Spring 2013 Metanode for File Metadata (1/2) •  Each node independently allocates disk space for the data blocks it writes •  Elected dynamically with the help of token server •  The new data block addresses are sent to the metanode •  When a node first accesses a file it: –  Periodically...
View Full Document

This note was uploaded on 02/11/2014 for the course CS 480 taught by Professor Staff during the Spring '08 term at Colorado State.

Ask a homework question - tutors are online