Difference between revisions of "User:StefanoZacchiroli/Content deduplication"

From Software Heritage Wiki
Jump to: navigation, search
Line 1: Line 1:
Some experiments on deduplicating contents at sub-file granularity using [https://en.wikipedia.org/wiki/Rabin_fingerprint Rabin fingerprints].
+
Some experiments on deduplicating contents at sub-file granularity.
  
 
== Linux kernel, Git repo ==
 
== Linux kernel, Git repo ==
Line 6: Line 6:
 
* 1.653.941 content blobs, for a total of 19 GB (compressed)
 
* 1.653.941 content blobs, for a total of 19 GB (compressed)
  
== test 1 ==
+
== Rabin fingerprints ==
 +
 
 +
Use [https://en.wikipedia.org/wiki/Rabin_fingerprint Rabin fingerprints].
 +
 
 +
=== test 1 ===
  
 
Rabin fingerprint parameters:
 
Rabin fingerprint parameters:
Line 18: Line 22:
 
* original size (uncompressed): 55.89 GB
 
* original size (uncompressed): 55.89 GB
 
* dedup chunk size (uncompressed): 19.87 GB (35.55%)
 
* dedup chunk size (uncompressed): 19.87 GB (35.55%)
 +
 +
== References ==
 +
 +
* [https://pdos.csail.mit.edu/papers/lbfs:sosp01/lbfs.pdf LBFS]
 +
* [http://www2009.eprints.org/7/1/p61.pdf Hailstorm]
 +
* [https://pdfs.semanticscholar.org/5231/82a9a66f96241d4bab0b92cc92e9cbda22c7.pdf DCT]
 +
* [https://github.com/nexB/scancode-toolkit/blob/ba5be09e54f57bd519239fd81746ebc4d0f9af9c/src/licensedcode/tokenize.py#L188 ScanCode tokenizer]

Revision as of 12:53, 7 January 2018

Some experiments on deduplicating contents at sub-file granularity.

Linux kernel, Git repo

  • origin: git.kernel.org, on 2018-01-06
  • 1.653.941 content blobs, for a total of 19 GB (compressed)

Rabin fingerprints

Use Rabin fingerprints.

test 1

Rabin fingerprint parameters:

  • prime: 3
  • window_size: 48 KB
  • min_block_size: 2 KB
  • avg_block_size: 8 KB
  • max_block_size: 64 KB

Results:

  • original size (uncompressed): 55.89 GB
  • dedup chunk size (uncompressed): 19.87 GB (35.55%)

References