Skip to content

dmeister/fs-c

Repository files navigation

Introduction

The fs-c tools can be used in two modes (Trace and Parse). In the Trace mode directories of a 
file system are processed by chunking all files using the specified chunking methods. While 
a direct analysis of the chunks if possible, often the chunk data is stored in a trace file 
for later processing. These chunk files then can be analyzed in detail during the Parse mode.

Trace Details

The trace mode is started with fs-c trace and accepts the following parameters:

-f   --filename       Name of a directories to parse. The application accepts multiple -f options
-c   --chunker        Chunking method (cdcX, fixedX with X in {2,4,8,16,32}
-o   --output         Output file (optional). If no output is specified, the chunks are analyzed 
                      directly
-t   --threads        Number of concurrent threads (Default: 1)
     --silent         Reduced output
-l   --listing        File contains a listing of files (-f option does not contain a directory, 
                      but a file that contains a directory to parse in each line)
-p   --privacy        Privacy Mode (hashes the filename to avoid storing concrete filenames in trace 
                      files for privacy reasons)
     --digest-length  Length of the fingerprint (default: 20 Bytes for SHA-1)
     --digest-type    Fingerprint type (MD5, SHA-1, ....)
     --report         Time interval between progress report messages in seconds (default: 60 seconds)        
     --custom-handler any fully qualified Scala class name implementing the FileDataHandler trait.
                      The class path can be extended by user defined classed via the FSC_EXTRA_CLASSPATH
                      environment variable.    
                     
Parse Details

The parse mode is started with fs-c parse and accepts the following parameters:

-t   --type  Type of the analysis. Currently supported:
              - "simple", that displays deduplication ratios for each chunked file (similar to 
                     fs-c trace without output option),
              - "ir" calculation deduplication ratios for each file type and file size categories,
              - "tr" calculating the temporal redundancy between two traces to simulate a 
                     backup scenario
              - "harniks" uses Harnik's estimation method to estimate the deduplication ratios.
                     All estimates have at most an error of 1% (or NaN is printed out).
                     Note that harniks and ir use a different method to assign chunks to file types/file
                     sizes. To that values between ir and harniks may differ. The hadoop scripts use the same
                     calculation method as harniks.
              - any fully qualified Scala class name implementing the FileDataHandler trait.
                     The class path can be extended by user defined classed via the FSC_EXTRA_CLASSPATH environment
                     variable. 
-o   --output    Run name
-f   --filename  Trace filename


Import Details

The import mode is started with fs-c import and accepts the following parameters:

-f   --filename Filename of the trace file to import
-r   --report   Interval between progress reports (in seconds, Default: 60s, 0 = no report)
-o   --output   HDFS directory as import target


Validate Details

The validate mode is started with fs-c validate and accepts the following parameters:

-f   --filename Filename of the trace file to validate
-r   --report   Interval between progress reports (in seconds, Default: 60s, 0 = no report)

About

Application to measure and analyze file systems to find the internal and temporal redundancy for file-based chunking and fingerprint-based data de-duplication.

Resources

Stars

Watchers

Forks

Packages

No packages published