| 
|
Topic: log-structured file system
topics >
computer science >
operating system >
backtracking
Topic:
database transactions
Topic:
examples of file systems
Topic:
file cache
Topic:
logging data and events
Topic:
log-structured rollback-recovery
Topic:
read-only and write-once file systems
Topic:
resourceful, redundant systems for reliability
Topic:
reversible execution
Summary
A log-structured file system writes all data to a sequential log. An index allows reads to recover the corresponding files. In contrast, a normal file system uses an index to write updates to sequential files. A log-structured file system assumes that writes will predominate over uncached reads.The main weakness of a log-structured file system is that obsolete data needs to be deleted from log segments. A segment cleaner recovers unallocated segments for writing new logs. Except for small files without segment cleaning, the overall performance of a log-structured file system is similar to a normal file system. (cbb 4/98)
Subtopic: why log-structured?
Quote: a log-structured file system writes all new information to a sequential log with an index; writes are efficient Quote: log-structured file system: all writes appended to end with indexing information for reads; since caches make writes predominate [»roseM10_1991]
Subtopic: segment cleaner
Quote: with increasing memory caches, disk traffic will become dominated by writes [»roseM2_1992]
Subtopic: comparison
Quote: a log-structured file system needs a segment cleaner process to recover large extents of free space from heavily fragmented segments [»roseM2_1992] Quote: disk fragmentation slows down the BSD Fast File System by 5-15%; segment cleaner overhead slows down a log-structured system by 33% [»seltM1_1995]
Subtopic: update logs
Quote: compared the BSD log-structured file system with the BSD Fast File System [»seltM1_1995] Quote: log-structured files much better for create and delete of small files; better for writes to small files; otherwise similar to other file systems
Subtopic: avoiding redo logs
Quote: log-based rollback-recovery must log all nondeterministic events with enough information to replay the event; useful for applications that frequently interact with the outside world [»elnoEN9_2002] Quote: write changes of the file name table and leader pages to a redo log; recovery in about two seconds; log entry format [»hagmR_1987]
Subtopic: append-only files
Quote: implement transactions in 5 microsec using persistent memory, file cache, and fast, recoverable memory management; no redo log or system calls [»loweDE10_1997]
Quote: Google's scalable distributed file system was designed for frequent component failure and huge, append-only files
[»gherS10_2003]
Topic: backtracking (30 items)
Topic: database transactions (27 items)
Topic: examples of file systems (44 items)
Topic: file cache (23 items)
Topic: logging data and events (17 items)
Topic: log-structured rollback-recovery (13 items)
Topic: read-only and write-once file systems (8 items)
Topic: resourceful, redundant systems for reliability (38 items)
Topic: reversible execution
(20 items)