this will cause better computation speed for single- and multi-core;
there are enhancements that will speed up old and slow machines as well
as multi-core CPUs. Indexing of surrogates has been speed up
from 4000 PPM to over 20000 PPM on a simple dual core office computer.
Since the enhancements are mostly in core routines, the hack should also
speed up search performance.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6276 6c8d7289-2bf4-0310-a012-ef5d649a1542
- added a open-on-demand hack to heap files: when a heap file is
opened the first time, it is first scanned to get a key index
and then it is closed again. This will free up file pointers
in cases where a really large number of blob files are opened
upon initialization of ArrayStack objects. This should solve
also a problem reported in
http://forum.yacy-websuche.de/viewtopic.php?p=17191#p17191
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6267 6c8d7289-2bf4-0310-a012-ef5d649a1542
- check routine at the beginning of the import of table keys that check that all imported keys are well-formed. All records that have a ill-formed key are deleted. This is a hack and is not tested since I don't have bad data here to test with. If the effect is seen in the wild, please report in the forum.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6245 6c8d7289-2bf4-0310-a012-ef5d649a1542
- better control to the cache by using combined request-header and content access methods
- refactoring of many classes to comply to this new access method
- make shure that the cache is always written if something was loaded
- some redesign of the process how http response results are feeded into the new indexing queue
- introduction of a cache read policy:
* never use the cache
* use the cache if entry exist
* use the cache if the proxy freshness rule confirmes
* use only the cache and go never online
- added configuration options for the crawl profiles to use the new cache policies. There is not yet a input during crawl start to set the policy but this will be added in another step.
- set the default policies for the existing crawl profiles. If you want them to appear in your default profiles you must delete the crawl profiles database; othervise the policy is 'proxy freshness rule'
- enhanced some cache access methods in such a way that unnecessary retrievals are omitted (i.e. for size computation). That should reduce some IO but also a lot of CPU computation because sizes were computed after decompression of content after retrieval of the content from the disc.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6239 6c8d7289-2bf4-0310-a012-ef5d649a1542
- removed the plasma package. The name of that package came from a very early pre-version of YaCy, even before YaCy was named AnomicHTTPProxy. The Proxy project introduced search for cache contents using class files that had been developed during the plasma project. Information from 2002 about plasma can be found here:
http://web.archive.org/web/20020802110827/http://anomic.de/AnomicPlasma/index.html
We stil have one class that comes mostly unchanged from the plasma project, the Condenser class. But this is now part of the document package and all other classes in the plasma package can be assigned to other packages.
- cleaned up the http package: better structure of that class and clean isolation of server and client classes. The old HTCache becomes part of the client sub-package of http.
- because the plasmaSwitchboard is now part of the search package all servlets had to be touched to declare a different package source.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6232 6c8d7289-2bf4-0310-a012-ef5d649a1542
- The indexing queue was a historic data structure that was introduced at the very beginning at the project as a part of the switchboard organisation object structure. Without the indexing queue the switchboard queue becomes also superfluous. It has been removed as well.
- Removing the switchboard queue requires that all servlets are called without a opaque generic ('<?>'). That caused that all serlets had to be modified.
- Many servlets displayed the indexing queue or the size of that queue. In the past months the indexer was so fast that mostly the indexing queue appeared empty, so there was no use of it any more. Because the queue has been removed, the display in the servlets had also to be removed.
- The surrogate work task had been a part of the indexing queue control structure. Without the indexing queue the surrogates needed its own task management. That has been integrated here.
- Because the indexing queue had a special queue entry object and properties attached to this object, the propterties had to be moved to the queue entry object which is part of the new indexing queue withing the blocking queue, the Response Object. That object has now also the new properties of the removed indexing queue entry object.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6225 6c8d7289-2bf4-0310-a012-ef5d649a1542
- added new Stack class that implements a stack on BLOB files
- added new Stacks class that can be used for a set of Stacks (a 'Stack Database')
- added methods to other classes to support the new stacks
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6169 6c8d7289-2bf4-0310-a012-ef5d649a1542
- removed migration code
- removed BLOBTree
after the removal of the BLOBTree, a lot of dead code appeared:
- removed dead code that was needed for BLOBTree
Some more classes may have not much use any more after the removal of BLOBTree, but still have some component that are needed elsewhere. Additional Refactoring steps are needed to clean up dependencies and then more code may appear that is unused and can be removed as well.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6150 6c8d7289-2bf4-0310-a012-ef5d649a1542
high-performance search query situations as seen in yacy-metager integration showed deadlock situation caused by synchronization effects inside of sun.java code. It appears that the logger is not completely safe against deadlock situations in concurrent calls of the logger. One possible solution would be a outside-synchronization with 'synchronized' statements, but that would further apply blocking on all high-efficient methods that call the logger. It is much better to do a non-blocking hand-over of logging lines and work off log entries with a concurrent log writer. This also disconnects IO operations from logging, which can also cause IO operation when a log is written to a file. This commit not only moves the logger from kelondro to yacy.logging, it also inserts the concurrency methods to realize non-blocking logging.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6078 6c8d7289-2bf4-0310-a012-ef5d649a1542
- added a analysis method that counts bytes that could be saved in case the new HandleMap can be applied in the most efficient way. Look for the log messages beginning with "HeapReader saturation": in most cases we could save about 30% RAM!
- removed the old FlexTable database structure. It was not used any more.
- removed memory statistics in PerformanceMemory about flex tables and node caches (node caches were used by Tree Tables, which are also not used any more)
- add a stub for a steering of navigation functions. That should help to switch off naviagtion computation in cases where it is not demanded by a client
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6034 6c8d7289-2bf4-0310-a012-ef5d649a1542
iterate them, then merge; not collect them and merge then.
This should use less memory and may behave better in an environment with many queries.
To ensure that too many queries will not cause total blocking,
a time-out of one second was also added. After the time-out
the index data that was collected so far is returned.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6013 6c8d7289-2bf4-0310-a012-ef5d649a1542
- fixed too early computation of navigation
- moved navigation rendering to yacysearchtrailer
- added more asserts
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@6006 6c8d7289-2bf4-0310-a012-ef5d649a1542
divided that class into three parts:
- the peers object is now hosted by the plasmaSwitchboard
- the crawler elements are now in a new class, crawler.CrawlerSwitchboard
- the index elements are core of the new segment data structure, which is a bundle of different indexes for the full text and (in the future) navigation indexes and the metadata store. The new class is now in kelondro.text.Segment
The refactoring is inspired by the roadmap to create index segments, the option to host different indexes on one peer.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5990 6c8d7289-2bf4-0310-a012-ef5d649a1542
This removes the last very IO-intensive data structures which were still used for Wiki, Blog and Bookmarks. Old database files will still remain in the DATA subdirectory but can be deleted manually if no major bugs appear during migration. There is no need for any user action, all migration is done automatically.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5986 6c8d7289-2bf4-0310-a012-ef5d649a1542
Other file sizes can now be configured with the attributes
filesize.max.win and filesize.max.other
the default maximum file size for non-windows OS is now 32GB
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5974 6c8d7289-2bf4-0310-a012-ef5d649a1542
Existing indexes based on the old index collections must be migrated with YaCy 0.8
- removed index collection classes and all migration tools
- added a 'incell' reference collection feature in URL analysis
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5966 6c8d7289-2bf4-0310-a012-ef5d649a1542
- increased default memory size to 180MB
- fixed possible bug in http client reset (there was a deadlock)
- bug in BOBHeap marked, but not solved, cause is still unknown.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5912 6c8d7289-2bf4-0310-a012-ef5d649a1542
orbiter