I have been investigating all the numerous disparate reasons that routingTime gets very, very high, causing the node to think it is overloaded. For example: Time Mean Std. Dev. Minimum Maximum Observations 2/26/03 12:49:00 AM GMT 1293.857142857143 1978.1820766985602 13.0 5643.0 21 2/26/03 12:50:00 AM GMT 96.86666666666666 81.85951961977881 15.0 296.0 15 2/26/03 12:51:00 AM GMT 235.65384615384616 284.3282337277129 16.0 1218.0 52 2/26/03 12:52:00 AM GMT 215.89583333333334 779.3465959237996 13.0 5421.0 48 .... 2/26/03 12:58:00 AM GMT 129.475 616.0853249056518 11.0 3925.0 40 2/26/03 12:59:00 AM GMT 53.770833333333336 89.68271248373188 13.0 624.0 48 2/26/03 1:00:00 AM GMT 821.1111111111111 2771.189464066222 17.0 11899.0 18
This happens on both Kaffe and Sun 1.4.1. According to my research it seems to all come down to lock contention on the RoutingTable. This was always going to be a problem, since every method in RoutingTable is synchronized. Hopefully I can implement a reader/writer lock on the actual tree, and lock the individual CPAlgoData's where necessary. -- Matthew Toseland [EMAIL PROTECTED]/[EMAIL PROTECTED] Full time freenet hacker. http://freenetproject.org/ Freenet Distribution Node (temporary) at http://80-192-4-23.cable.ubr09.na.blueyonder.co.uk:8889/K4sKgHdHRUw/ ICTHUS.
pgp00000.pgp
Description: PGP signature
