Re: [freenet-support] Freenet 0.7.5 build 1327 (small, preparatory)
On Sun, 16 Jan 2011 10:06:04 +0300, Volodya wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 01/16/2011 08:33 AM, Dennis Nezic wrote: On Sun, 16 Jan 2011 08:00:08 +0300, Volodya wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 (Off topic hand waving, how hard is it, honestly, to have a counter on each incoming packet, and to only accept inputBandwidthLimit per second? (And perhaps to add clever debugging messages, for now, if the node is ever seen significantly exceeding this limit?)) How do you count the bandwidth? Packets are discrete while bandwidth is continuous. Let's say that you have a bandwidth limit of 5 KiB per second; now imagine that a packet comes in that is 42 kilobits, if you download that packet in less than a second, you have gone over your limit. But you cannot slow down half way through a packet. You have said that you do not think that Freenet should average things out, but that is the only way to move from discrete to continuous math while preserving any sense of sanity. True. Are Freenet's packets that big? (Smaller packet sizes would be one solution for better bandwidth management.) I guess the discrete spikes aren't really a problem though, but rather those massive and continuous flows that swamp my connection -- as if the packet counter suddenly decided to stop working altogether, or something. I have just given an example, it had little to do with Freenet packet size. Still what i was pointing towards is that you need to average things out over some period of time. Perhaps the problem is in something different, but it could be that it selects some time period that is too large or too small. (In the meantime, till the bug (and possibly the feature) settle down, I have put in place an iptables rule+tc bandwidth limiting filter.) ___ Support mailing list Support@freenetproject.org http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:support-requ...@freenetproject.org?subject=unsubscribe
Re: [freenet-support] Freenet 0.7.5 build 1327 (small, preparatory)
On Thu, 13 Jan 2011 10:21:37 -0500, Dennis Nezic wrote: On Wed, 12 Jan 2011 20:31:31 +, Matthew Toseland wrote: Build 1327 includes three bugfixes which will help to debug 1328, which hopefully will finally include the new load management prerequisites (sending the load messages, two-level timeouts). It will be mandatory on Monday but please upgrade now. The three fixes are related to: - Even if a request takes too long and times out, we need to tell the previous node about the timeout. This is necessary for two-stage timeout in the next build, which will look out for it to tell whether the timeout is the fault of the node we are waiting for or is a downstream error we can't do anything about. - Ignore some errors caused by nodes still running the old packet format (we work around these and they are not interesting; the same error is however interesting if it happens with new packet format). - Regard a message as sent after everything in the message has been sent once, even if it's not all in flight now due to packets being lost causing resends (sent is distinct from acknowledged). THANKS! Please report any problems you find! 1327 seems a lot more obedient, and hasn't gone berserk after about a day of uptime. (Muchos Gracias.) 1327 also goes berserk sometimes, although perhaps less often. Oddly enough, it goes way way over my limit (for at least a few minutes) when my connection is congested from other activity. (Aka. the worst possible time to go berserk.) (Off topic hand waving, how hard is it, honestly, to have a counter on each incoming packet, and to only accept inputBandwidthLimit per second? (And perhaps to add clever debugging messages, for now, if the node is ever seen significantly exceeding this limit?)) ___ Support mailing list Support@freenetproject.org http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:support-requ...@freenetproject.org?subject=unsubscribe
Re: [freenet-support] Freenet 0.7.5 build 1327 (small, preparatory)
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 (Off topic hand waving, how hard is it, honestly, to have a counter on each incoming packet, and to only accept inputBandwidthLimit per second? (And perhaps to add clever debugging messages, for now, if the node is ever seen significantly exceeding this limit?)) How do you count the bandwidth? Packets are discrete while bandwidth is continuous. Let's say that you have a bandwidth limit of 5 KiB per second; now imagine that a packet comes in that is 42 kilobits, if you download that packet in less than a second, you have gone over your limit. But you cannot slow down half way through a packet. You have said that you do not think that Freenet should average things out, but that is the only way to move from discrete to continuous math while preserving any sense of sanity. - Volodya - -- http://freedom.libsyn.com/ Echo of Freedom, Radical Podcast None of us are free until all of us are free.~ Mihail Bakunin -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.10 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iQEcBAEBAgAGBQJNMntRAAoJENW9VI+wmYasYEMH/1PK+Ea33qJZRGDEvFbktznX MRzyXFT/GYH3KTPQvD/aHcKcN81Rtc3aVCY1sHJzkdb2SmeSoV1Kilyd/Iw2kcl5 lk4HNd+to7XBQcsggBBPYMMV5I/GGQDdI1DYAbOmkcU+B7oTElCEktsI4PcyGNbk 7lBViMjqFYnhZhQDM3dqsJv6ik5WqHIS5DY3OD8b/C3ngERm3Au9pFBw8HGWwve9 stSrFhideHkjpSaaeoJMEt1L0cEDH42Mr6LO4eO1UdLnM8aG3khUXxvrPxylN8Kb WjI27kJc3DMnCefgL2xrtahjUVgTjBIJUgmTLIDLY3dFcnhlQl0C4jHfQf2QGjk= =eYnE -END PGP SIGNATURE- ___ Support mailing list Support@freenetproject.org http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:support-requ...@freenetproject.org?subject=unsubscribe
Re: [freenet-support] Freenet 0.7.5 build 1327 (small, preparatory)
On Sun, 16 Jan 2011 08:00:08 +0300, Volodya wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 (Off topic hand waving, how hard is it, honestly, to have a counter on each incoming packet, and to only accept inputBandwidthLimit per second? (And perhaps to add clever debugging messages, for now, if the node is ever seen significantly exceeding this limit?)) How do you count the bandwidth? Packets are discrete while bandwidth is continuous. Let's say that you have a bandwidth limit of 5 KiB per second; now imagine that a packet comes in that is 42 kilobits, if you download that packet in less than a second, you have gone over your limit. But you cannot slow down half way through a packet. You have said that you do not think that Freenet should average things out, but that is the only way to move from discrete to continuous math while preserving any sense of sanity. True. Are Freenet's packets that big? (Smaller packet sizes would be one solution for better bandwidth management.) I guess the discrete spikes aren't really a problem though, but rather those massive and continuous flows that swamp my connection -- as if the packet counter suddenly decided to stop working altogether, or something. ___ Support mailing list Support@freenetproject.org http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:support-requ...@freenetproject.org?subject=unsubscribe
Re: [freenet-support] Freenet 0.7.5 build 1327 (small, preparatory)
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 01/16/2011 08:33 AM, Dennis Nezic wrote: On Sun, 16 Jan 2011 08:00:08 +0300, Volodya wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 (Off topic hand waving, how hard is it, honestly, to have a counter on each incoming packet, and to only accept inputBandwidthLimit per second? (And perhaps to add clever debugging messages, for now, if the node is ever seen significantly exceeding this limit?)) How do you count the bandwidth? Packets are discrete while bandwidth is continuous. Let's say that you have a bandwidth limit of 5 KiB per second; now imagine that a packet comes in that is 42 kilobits, if you download that packet in less than a second, you have gone over your limit. But you cannot slow down half way through a packet. You have said that you do not think that Freenet should average things out, but that is the only way to move from discrete to continuous math while preserving any sense of sanity. True. Are Freenet's packets that big? (Smaller packet sizes would be one solution for better bandwidth management.) I guess the discrete spikes aren't really a problem though, but rather those massive and continuous flows that swamp my connection -- as if the packet counter suddenly decided to stop working altogether, or something. I have just given an example, it had little to do with Freenet packet size. Still what i was pointing towards is that you need to average things out over some period of time. Perhaps the problem is in something different, but it could be that it selects some time period that is too large or too small. - Volodya - -- http://freedom.libsyn.com/ Echo of Freedom, Radical Podcast None of us are free until all of us are free.~ Mihail Bakunin -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.10 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iQEcBAEBAgAGBQJNMpjAAAoJENW9VI+wmYasuf4H/0qUPAfEEYIKAWfdaweVlZCv dqTShmmNur1eQ+96ZWxTI0qbQ1+vYe774l25m0dAK9uds4nK+v0G+duDRhUdrkMY wSwOsfDXMqkyg7WgI4D9IIzer0nJoUJAfYhjrpwRA7fhX3OQJQizyMdEhL34bc4f 3kUe6ItH1A3ZYdpxdrp3TsRiT2hiciWimmqV2+D+P1aMK5ZBVxAPdY7qUgA1UpOh MSAhmfSLfQQOJrbdwejfO5eftNhYgMyrAuYPXCAkb3GNL1IPTE54eQ0BcMlPOpmj jfp1IgrHHDmNP7fVDOn7goxf5Vn/v3rIVdHXTTgbyUlh46EFC7h4/J6Ka+gKa3E= =WaFe -END PGP SIGNATURE- ___ Support mailing list Support@freenetproject.org http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:support-requ...@freenetproject.org?subject=unsubscribe
Re: [freenet-support] Freenet 0.7.5 build 1327 (small, preparatory)
On Wed, 12 Jan 2011 20:31:31 +, Matthew Toseland wrote: Build 1327 includes three bugfixes which will help to debug 1328, which hopefully will finally include the new load management prerequisites (sending the load messages, two-level timeouts). It will be mandatory on Monday but please upgrade now. The three fixes are related to: - Even if a request takes too long and times out, we need to tell the previous node about the timeout. This is necessary for two-stage timeout in the next build, which will look out for it to tell whether the timeout is the fault of the node we are waiting for or is a downstream error we can't do anything about. - Ignore some errors caused by nodes still running the old packet format (we work around these and they are not interesting; the same error is however interesting if it happens with new packet format). - Regard a message as sent after everything in the message has been sent once, even if it's not all in flight now due to packets being lost causing resends (sent is distinct from acknowledged). THANKS! Please report any problems you find! 1327 seems a lot more obedient, and hasn't gone berserk after about a day of uptime. (Muchos Gracias.) ___ Support mailing list Support@freenetproject.org http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:support-requ...@freenetproject.org?subject=unsubscribe
[freenet-support] Freenet 0.7.5 build 1327 (small, preparatory)
Build 1327 includes three bugfixes which will help to debug 1328, which hopefully will finally include the new load management prerequisites (sending the load messages, two-level timeouts). It will be mandatory on Monday but please upgrade now. The three fixes are related to: - Even if a request takes too long and times out, we need to tell the previous node about the timeout. This is necessary for two-stage timeout in the next build, which will look out for it to tell whether the timeout is the fault of the node we are waiting for or is a downstream error we can't do anything about. - Ignore some errors caused by nodes still running the old packet format (we work around these and they are not interesting; the same error is however interesting if it happens with new packet format). - Regard a message as sent after everything in the message has been sent once, even if it's not all in flight now due to packets being lost causing resends (sent is distinct from acknowledged). THANKS! Please report any problems you find! signature.asc Description: This is a digitally signed message part. ___ Support mailing list Support@freenetproject.org http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:support-requ...@freenetproject.org?subject=unsubscribe