because of the strongly enhanced indexing speed when using the new IndexCell RWI data structures (> 2000PPM on my notebook), it is now necessary to control the crawling speed depending on the response time of the target server (which is also YaCy in case of some intranet indexing use cases).
The latency factor in crawl delay times is derived from the time that a target hosts takes to answer on http requests. For internet domains, the crawl delay is a minimum of twice the response time, in intranet cases the delay time is now a halve of the response time.
- added API to monitor the latency times of the crawler:
a new api at /api/latency_p.xml returns the current response times of domains, the time when the domain was accessed by the crawler the last time and many more attributes.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5733 6c8d7289-2bf4-0310-a012-ef5d649a1542
this option was never used and there is also no use to set other columns but the first as the primary key. as a result, access methods to the key do not need to compute key positions, and they work faster.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5711 6c8d7289-2bf4-0310-a012-ef5d649a1542
The speed of the kelondro indexing class ObjectIndexCache can be compared with Javas standard TreeMap with the main method in IntegerHandleIndex. The result is, that the kelondro indexing needs only 1/5 of the memory that TreeMap uses! In exchange, the kelondro classes are slower than TreeMap, about four (!) times slower. However, this is not so bad because the better use of the memory is a strong advantage and makes it possible that YaCy can maintain such a large number of document (> 50 million) in one peer.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5705 6c8d7289-2bf4-0310-a012-ef5d649a1542
without an order by the primary key. The result is a very fast enumeration of the Eco table data structure. Other table data types are not affected.
The new enumerator is used for the url export function that can be accessed from the online interface (Index Administration -> URL References -> Export). This export should now be much faster, if all url database files are from type Eco
The new enumeration is also used at other functions in YaCy, i.e. the initialization of the crawl balancer and the initialization of YaCy News.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5647 6c8d7289-2bf4-0310-a012-ef5d649a1542
- after a index selection is made, the index is splitted into its vertical components
- from differrent index selctions the splitted components can be accumulated before they are placed into the transmission queue
- each splitted chunk gets its own transmission thread
- multiple transmission threads are started concurrently
- the process can be monitored with the blocking queue servlet
To implement that, a new package de.anomic.yacy.dht was created. Some old files have been removed.
The new index distribution model using a vertical DHT was implemented. An abstraction of this model
is implemented in the new dht package as interface. The freeworld network has now a configuration
of two vertial partitions; sixteen partitions are planned and will be configured if the process is bug-free.
This modification has three main targets:
- enhance the DHT transmission speed
- with a vertical DHT, a search will speed up. With two partitions, two times. With sixteen, sixteen times.
- the vertical DHT will apply a semi-dht for URLs, and peers will receive a fraction of the overall URLs they received before.
with two partitions, the fractions will be halve. With sixteen partitions, a 1/16 of the previous number of URLs.
BE CAREFULL, THIS IS A MAJOR CODE CHANGE, POSSIBLY FULL OF BUGS AND HARMFUL THINGS.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5586 6c8d7289-2bf4-0310-a012-ef5d649a1542
- ensuring that ordered indexes stay ordered during remove
- no unnecessary ordering checks
- better test logic in crawl stacker
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5457 6c8d7289-2bf4-0310-a012-ef5d649a1542
The cache is now flushed only for one second every ten seconds. During a crawl the cache
fills up completely, and is only flushed if space is needed for more documents.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5446 6c8d7289-2bf4-0310-a012-ef5d649a1542
- introduced blocking queues in CrawlStacker to make it ready for concurrency
- added a second busy thread for the CrawlStacker
The CrawlStacker is multithreaded. It shall be transformed into a BlockingThread in another step.
The concurrency of the stacker will hopefully solve some problems with cases where DNS blocks.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5395 6c8d7289-2bf4-0310-a012-ef5d649a1542
- added a write buffer to BLOBHeap
- modified the BLOBBuffer (is now only to buffer non-compressed content)
- added content compression to the HTCache
The new read cache will decrease the start/initialization time of BLOB files,
like the HTCache, RobotsTxt and other BLOBHeap structures.
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5386 6c8d7289-2bf4-0310-a012-ef5d649a1542
orbiter