Kenneth Skovhede
c34023139f
Refactored the database to store the remote volume id instead of the filename for each block, which reduces required storage space for the local database
2013-03-22 21:56:00 +01:00
Kenneth Skovhede
ed7ed5e6c2
Fixed a bug that caused 8x larger memory to be used than what was specified
2013-03-16 12:30:29 +01:00
Kenneth Skovhede
0a80b04352
Made it possible to deactivate all cache tables
2013-03-16 10:37:57 +01:00
Kenneth Skovhede
6d2e39bb38
Added support code to work with the file-change-monitor
2013-03-15 22:04:07 +01:00
Kenneth Skovhede
7bc6c2b6ff
Rewrote the way the restore operation uses file lists for partial restore
2013-03-15 21:16:29 +01:00
Kenneth Skovhede
ac90565c39
Cleaned up some unused code
2013-03-15 21:15:51 +01:00
Kenneth Skovhede
f6a8511d77
Refactored the hash lookup.
...
Added hash lookup for metadata.
Added optional path-based lookup, which are disabled because they use a lot of memory.
2013-03-15 21:14:00 +01:00
Kenneth Skovhede
1231dc24da
Made the hash memory requirement configurable
2013-03-11 20:45:48 +01:00
Kenneth Skovhede
b7531b3766
Added double lookup for block and file hash, this reduces the execution time to half.
2013-03-11 20:13:40 +01:00
Kenneth Skovhede
11e863eb78
Fixed a bug with the hash lookup table, much faster now
2013-03-10 23:01:17 +01:00
Kenneth Skovhede
f055dfd2b4
Optimized block queries
2013-03-10 20:33:30 +01:00
Kenneth Skovhede
5863292f11
Added hash prefix lookup to reduce database pressure
2013-03-10 14:27:13 +01:00
Kenneth Skovhede
f6d963ddb9
Added extra convenience method
2013-03-10 14:25:45 +01:00
Kenneth Skovhede
513fc9b3d2
Reduced a bottleneck related to storing metadata, and added an option to avoid storing metadata
2013-03-09 13:25:50 +01:00
Kenneth Skovhede
aca4ed95cc
More work on the block hash.
...
Rewrote the backup and restore to work within a transaction as that speeds things up a lot.
Rewrote the backup procedure to generate the file list from the database, instead of on-the-fly as that is more efficient.
Re-arranged the database to use less storage on large sets of unchanged files.
2013-03-08 22:24:54 +01:00
kenneth@hexad.dk
a0521da54d
Initial version of the new block-based storage format.
...
Still needs a work over in terms of error handling, progress reporting, warning output, refactoring etc.
Also misses the volume reuse/reclaim algorithms.
But it passes the unit tests now, with and without a local database.
git-svn-id: https://duplicati.googlecode.com/svn/sandboxes/Kenneth/ForestHash@1526 59da171f-624f-0410-aa54-27559c288bec
2013-02-12 21:43:14 +00:00
Kenneth Skovhede