Re: [freenet-support] Way to much RAM! Build 5064
I've repeatedly seen old machines like my P3-600 disregarded as irrelevant, and not worth optimizing for, in terms of the Freenet network. See above. The best thing I can do for you is get rate limiting working properly. And I think Freenet should easily run on a 600MHz machine, or something is wrong. I'm just skeptical about running on 128MB, or on 200MHz machines. I dont think it impossible to run a node on 128MB.. However.. if the OS uses up 80 of those we will definitely end up having problems. /N ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
[freenet-support] freenet 5069 bug report - upload files nearly impossible (FEC)
I'm not able to upload files ( 5MB) into stable-network. Details Since some weeks I try to upload some healing blocks into stable network, on average the FEC-block size is 256 KB. Normally I use FUQID 1.3, but because it transfered 0.1 KB/s and reported a lot of Insert thread failed. Retrying... in the log, I also tried it directly with FProxy with the same result. When I set the Insertion-Thread-Number to a higher value than 5 the Current messageSendTimeRequest value in General Information of the web interface gets greater than 200%-300%. The maximum was 1%-15000% when I set the thread number to 30. The Transfers active value grows also from ~20-30 to ~180. Fproxy mostly reported ROUTE NOT FOUND on blocks where inserting was failed. Also the Instantaneous local traffic decreases significately. Heal HTL: 15 delete key from local node vefore inserting is checked The upstream bandwith usage is ~ 1 - 12500 B/sec. (12000 Maximum) Typically my node is running for periods of 3 - 10 h a day. Plattform: WinXP Pro JVM: 1.4.2_03-b02 Freenet: Fred,0.5,STABLE-1.50,5069,CVS:1.90.2.50.2.96 Detailed System Information: Plattform: WinXP Pro, Pentium 3, 933 MHz, 512 MB RAM JVM: Sun, 1.4.2_03-b02 Freenet: Fred,0.5,STABLE-1.50,5069,CVS:1.90.2.50.2.96 Internet Connection: DSL: 768 kbit downstream / 128 kbit upstream Data Store: 7 GiB, Percent used: 95 Mean Ustream Traffic: 1 - 12500 bytes/second unstable.ref: 2004-02-15 Typically number of connections: 100 - 180 diff default.ini/freenet.ini: ipAddress=XXX listenPort=XXX fcpHosts=127.0.0.1,localhost doAnnounce=yes storeSize=7168M inputBandwidthLimit=7 outputBandwidthLimit=12000 logLevel=Normal rtMaxRefs=51 rtMaxNodes=51 maxHopsToLive=25 maximumThreads=130 tempDir=XXX mainport.params.servlet.1.params.passThroughMimeTypes=text/plain,image/jpeg,image/gif,image/png Regards Rudi # Some Statistics # # Node ran 1 h before I started an upload with 30 Threads # now it runs ~3 h # -- Routing Table status: 18.02.2004 11:27:28 Number of node references 51 Attempted to contact node references17 Contacted node references 48 Connections with Successful Transfers 21 Backed off nodes17 Connection Attempts 1646 Successful Connections 1527 Lowest max estimated search time0ms Lowest max estimated DNF time 0ms Lowest global search time estimate 289962ms Highest global search time estimate 460245ms Lowest global transfer rate estimate2.318 bytes/second Highest global transfer rate estimate 3.643 bytes/second Lowest one hop probability of DNF 0,98 Highest one hop probability of DNF 0,99 Lowest one hop probability of transfer failure 0,79 Highest one hop probability of transfer failure 0,92 Single hop probability of QueryRejected 0.29645163211077075 Single hop average time for QueryRejected 87863.09895443736 Single hop probability of early timeout 0.6783154528702626 Single hop average time for early timeout 24009.760728425113 Single hop probability of search timeout0.9186481377761555 Single hop average time for search timeout 316643.7063678293 Total number of requests that didn't QR 23767 Total number of reqests that timed out before a QR or Accepted 6676 Implementation freenet.node.rt.NGRoutingTable #... # #-- Load # Current routingTime 0ms Current messageSendTimeRequest 7220ms Pooled threads running jobs 5 (3,8%) Pooled threads which are idle 25 Current upstream bandwidth usage11215 bytes/second (93,5%) Reason for refusing connections:avgMessageSendTimeRequest(7220,250) successfulSendTimeCutoff(2000,000) Reason for QueryRejecting requests: Estimated load (100%) overloadHigh (80%)Estimated load (100%) overloadHigh (80%) It's normal for the node to sometimes reject connections or requests for a limited period. If you're seeing rejections continuously the node is overloaded or something is wrong (i.e. a bug). Current estimated load for QueryReject purposes 100% Current estimated load for rate limiting722% [Rejecting incoming connections and requests!] Reason for load:Load due to thread limit = 3,8% Load due to routingTime = 0% = 0% / 100.000% = overloadLow (50%) Load due to messageSendTimeRequest = 722% = 722.025% / 100.000% overloadLow (50%) Load due to output bandwidth limiting = 93,5% because outputBytes(672911) limit (576000,009 ) = outLimitCutoff (0,8) * outputBandwidthLimit (12000) * 60 Estimated external pSearchFailed (based only on QueryRejections due to load): 0.9998 Current estimated requests per hour (based on last 10 mins):5772.0 Current global quota (requests per hour): 799.4182883301385 Current recommended request interval sent to client nodes 2.147483647E9ms ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at
[freenet-support] Interesting observations regarding performance
Hi all, I think some will find this interesting. I've run Freenet on a Duron 900 with 512Mb ram, Windows 2000, for quite some time. The javaw process has consumed all available CPU, and the computer has been quite sluggish due to Freenet. I've also seen very low transfer speeds in FUQID (easiest to measure with, I think) - around 1-2kb/s when content is found. A few days ago I got a new system - a P4 3.2GHz, 1024Mb ram, still Windows 2000. Freenet now consumes around 0-1% CPU-time, it has no negative effect whatsoever on the performance of the rest of the system, and transfer speeds are up to 11kb/s. The Freenet node and contents are the same, I moved system behind my NAT so the Freenet network knows nothing of my change - it's all internal. I find the above very interesting .. comments from others? regards, Troed -- http://troed.se - controversial views or common sense? ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
[freenet-support] freenethelp.org domain
Hi all, Last July I registered the domain freenethelp.org with the intent of creating a web-based support site for Freenet. My goal was to provide tutorials for installing Freenet on various platforms, to answer FAQs, etc. as well as offer a message board. The idea was that finding help about Freenet _on_ Freenet is slow and cumbersome for a newbie, and documentation about Freenet on the official site is rather sparse. I decided that a normal non-anonymous web resource might be able to give new users a head start. Alas, more than six months later, I have yet to find the time to pursue this in any meaningful way. I am willing to donate the domain to the Freenet Project, or to anyone who has the time and resources to fulfill the goals I had when I registered the domain. Anyone who is serious about creating a help site for Freenet, please contact me on or off list ([EMAIL PROTECTED]). I will give the domain to The Freenet Project, no questions asked, if Ian or Toad would like to have it. Otherwise, anyone who's interested is eligible. -s ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
Re: [freenet-support] Interesting observations regarding performance
On Wed, Feb 18, 2004 at 12:36:01PM +0100, Troed S?ngberg wrote: Hi all, I think some will find this interesting. I've run Freenet on a Duron 900 with 512Mb ram, Windows 2000, for quite some time. The javaw process has consumed all available CPU, and the computer has been quite sluggish due to Freenet. I've also seen very low transfer speeds in FUQID (easiest to measure with, I think) - around 1-2kb/s when content is found. A few days ago I got a new system - a P4 3.2GHz, 1024Mb ram, still Windows 2000. Freenet now consumes around 0-1% CPU-time, it has no negative effect whatsoever on the performance of the rest of the system, and transfer speeds are up to 11kb/s. The Freenet node and contents are the same, I moved system behind my NAT so the Freenet network knows nothing of my change - it's all internal. I find the above very interesting .. comments from others? Memory usage on both? regards, Troed -- http://troed.se - controversial views or common sense? -- Matthew J Toseland - [EMAIL PROTECTED] Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so. signature.asc Description: Digital signature ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
Re: [freenet-support] Interesting observations regarding performance
I've run Freenet on a Duron 900 with 512Mb ram, Windows 2000, for quite some time. The javaw process has consumed all available CPU, and the computer has been quite sluggish due to Freenet. I've also seen very low transfer speeds in FUQID (easiest to measure with, I think) - around 1-2kb/s when content is found. A few days ago I got a new system - a P4 3.2GHz, 1024Mb ram, still Windows 2000. Freenet now consumes around 0-1% CPU-time, it has no negative effect whatsoever on the performance of the rest of the system, and transfer speeds are up to 11kb/s. The Freenet node and contents are the same, I moved system behind my NAT so the Freenet network knows nothing of my change - it's all internal. I find the above very interesting .. comments from others? The most likely reason is that Duron (as well as Celeron) lacks enough L2 cache to effectively cache interpreted (or semi-interpreted) code, such as Java, Perl or PHP. For our serverside Java applications (which, admittedly, load CPU much higher than Freenet does) difference with Duron and Athlon is about 2x-4x at the same clock speed. We have no benchmarks for our GUI, but I've made one w/ Sun Forte for Java about 2-3 years ago. On the same machine equipped with Celeron 400 MHz (128 Kb L2 Cache) and PII-400 (512 Kb L2 cache), 128 Mb of RAM, Forte loaded in about 8 mins w/ Celeron, less than 3 minutes w/ full PII. Of course, Java went quite far in the past couple of years, but I still doubt it can fit in small amount of L2 cache available on modern low-end CPUs. Regards, Victor Denisov, CEO, Jera Systems, Moscow, Russia ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
Re: [freenet-support] Interesting observations regarding performance
On Wed, 18 Feb 2004 13:42:48 +, Toad [EMAIL PROTECTED] wrote: Memory usage on both? Roughly the same at ~150Mb (javaw process) Knock down the xferrate a bit from 11kb/s, it dropped a while after I sent that mess (I have no good explanation) but it's still a bit higher than on the previous machine. What surprised me was how transparent Freenet suddenly became, the difference is huge - something I cannot really justify looking at how much memory the old machine had and the CPU clock difference being only 3.5x. regards, Troed -- http://troed.se - controversial views or common sense? ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
Re: [freenet-support] Interesting observations regarding performance
On Wed, Feb 18, 2004 at 03:36:52PM +0100, Troed S?ngberg wrote: On Wed, 18 Feb 2004 13:42:48 +, Toad [EMAIL PROTECTED] wrote: Memory usage on both? Roughly the same at ~150Mb (javaw process) Knock down the xferrate a bit from 11kb/s, it dropped a while after I sent that mess (I have no good explanation) but it's still a bit higher than on the previous machine. What surprised me was how transparent Freenet suddenly became, the difference is huge - something I cannot really justify looking at how much memory the old machine had and the CPU clock difference being only 3.5x. Well, it could easily be closer to 7x in actual performance - the Duron has a LOT less cache than the faster machine... regards, Troed -- http://troed.se - controversial views or common sense? ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED] -- Matthew J Toseland - [EMAIL PROTECTED] Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so. signature.asc Description: Digital signature ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
Re: [freenet-support] Re: routing table
Well there are Countrys where bandwidth is very cheap or free for home users while memory is really expensive. For around 10$ per month I get unlimited bandwitdh on my adsl line. Take a look at the attached Image to see how cheap brand memory can be (noname memory isn't a subject cause it is a shure way to make your system unstable). This is one of the most expensive here, but it is still a quite good example for way higher prices in different countrys. Wow!!! Man, how do you use computers at all with such prices for brand memory? I've just checked - retail Kingston 256 Mb PC133 module costs $38-$45 here in Russia, and we consider our computer parts market to be quite expensive compared to western one. On the other hand, our broadband access is really expensive - the best you can have here in Moscow is either 7500/768 with 5 Gb traffic limit (and $10/Gb over the limit) or 128/64 unlimited, both for $99/month. Regards, Victor Denisov, CEO, Jera Systems, Moscow, Russia ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
[freenet-support] Re: routing table
Victor Denisov schrieb: Wow!!! Man, how do you use computers at all with such prices for brand memory? I've just checked - retail Kingston 256 Mb PC133 module costs $38-$45 here in Russia, and we consider our computer parts market to be quite expensive compared to western one. Like I said this is one of the most expensive ones, you can get brand 256 MB PC133 memory for around 80 euro to 90 euro. The prices for PC133 are also higher than the prices for DDR modules because they did become really rare here. But DDR modules don't fit into the older machines most people use for server purposes. ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
[freenet-support] Re: Compilation error
Yoann schrieb: I try to compile the lastest Freenet sources. I take the library freenet-ext.jar and junit.jar And when i do make, I have (I had -depecation) : javac -target 1.1 -classpath lib/freenet-ext.jar:lib/junit.jar -sourcepath src -d build -deprecation src/freenet/client/*.java src/freenet/client/cli/*.java src/freenet/support/servlet/http/HttpSessionImpl.java:7: warning: javax.servlet.http.HttpSessionContext in javax.servlet.http has been deprecated import javax.servlet.http.HttpSessionContext; ^ 1 warning I don't know what it mean This means that the named class/interface is from an older JVM version and shouldn't be used anymore because it is considered bad in some ways. But it isn't an error, it's just a warning and not really harmfull. ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
Re: [freenet-support] Compilation error
Le mer 18/02/2004 à 17:44, Toad a écrit : I don't know what it mean It means you should turn off -deprecation if you don't want to see the message :) If I turn off -deprecation, I have this : javac -target 1.1 -classpath lib/freenet-ext.jar:lib/junit.jar -sourcepath src -d build src/freenet/client/*.java src/freenet/client/cli/*.java Note: src/freenet/support/servlet/http/HttpSessionImpl.java uses or overrides a deprecated API. Note: Recompile with -deprecation for details. .jar are not generated :( -- Offset signature.asc Description: Ceci est une partie de message =?ISO-8859-1?Q?num=E9riquement?= =?ISO-8859-1?Q?_sign=E9e=2E?= ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
[freenet-support] Re: Compilation error
Yoann schrieb: If I turn off -deprecation, I have this : javac -target 1.1 -classpath lib/freenet-ext.jar:lib/junit.jar -sourcepath src -d build src/freenet/client/*.java src/freenet/client/cli/*.java Note: src/freenet/support/servlet/http/HttpSessionImpl.java uses or overrides a deprecated API. Note: Recompile with -deprecation for details. .jar are not generated :( How do you build it and on what OS? If you build it on windows use ant ( http://ant.apache.org/ ) the included build.bat doesn't work right. On linux the make file should work. ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
Re: [freenet-support] Compilation error
On Wed, Feb 18, 2004 at 05:36:44PM +0100, Yoann wrote: I try to compile the lastest Freenet sources. I take the library freenet-ext.jar and junit.jar And when i do make, I have (I had -depecation) : javac -target 1.1 -classpath lib/freenet-ext.jar:lib/junit.jar -sourcepath src -d build -deprecation src/freenet/client/*.java src/freenet/client/cli/*.java src/freenet/support/servlet/http/HttpSessionImpl.java:7: warning: javax.servlet.http.HttpSessionContext in javax.servlet.http has been deprecated import javax.servlet.http.HttpSessionContext; ^ 1 warning I don't know what it mean It means you should turn off -deprecation if you don't want to see the message :) Seriously though, this is annoying :| -- Matthew J Toseland - [EMAIL PROTECTED] Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so. signature.asc Description: Digital signature ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
Re: [freenet-dev] Re: [freenet-support] Compilation error
javac -target 1.1 -classpath lib/freenet-ext.jar:lib/junit.jar -sourcepath src -d build -deprecation src/freenet/client/*.java src/freenet/client/cli/*.java src/freenet/support/servlet/http/HttpSessionImpl.java:7: warning: javax.servlet.http.HttpSessionContext in javax.servlet.http has been deprecated import javax.servlet.http.HttpSessionContext; ^ 1 warning I don't know what it mean It means you should turn off -deprecation if you don't want to see the message :) Seriously though, this is annoying :| This is the last and only deprecation warning. Unfortunately it cannot be eliminated because, according to the Servlet 2.2 specification, getSession() must still exists in the HttpServletRequest class. As long as getSession() exists then the HttpSession class must exist, which means the deprecated class HttpSessionContext must exist. It's a little ugly, but it's harmless, although it may point out we are reinventing the wheel and maybe should be borrowing someone elses servlet implementation. ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
[freenet-support] Just in case...here's the complete set of latest scripts
[ I was going to send this only to Toad, but then decided it may be good to do one more post containing the complete set of the latest script revisions, all in one place. Hope no one minds. I tar/gzipped them all to save space. :-) ] Just in case you may have overlooked it, I recently posted a whole set of updated scripts, including (finally!) the dedupe script you [Toad] were asking for. I incorporated the deduping into the addnodes.sh script as well (the script that fetches noderefs.txt and updates seednodes.ref from it). It's all working very well here. I now have a completely deduped seednodes.ref file that is up to 203 unique nodes, with a decided bias for newer versions as well. :-) $ ./versions.sh seednodes.ref 7 version=Fred,0.6,1.50,6481 12 version=Fred,0.6,1.50,6482 67 version=Fred,0.6,1.50,6483 1 version=Fred,0.6,1.50,6484 2 version=Fred,0.6,1.50,6485 114 version=Fred,0.6,1.50,6486 The addnodes.sh script, by the way, is designed to always favor the latest nodes in noderefs.txt over any existing ones in seednodes.ref, so the resulting seednodes.ref will, over time, lean more and more to newer, currently active versions of nodes. You can run dedupe.sh on any noderefs file you like, by the way. Hint: if you cat a newer nodes file (say, the latest unstable.ref), followed by an older nodes file, saving to a new concatenated file, and then dedupe the new, concatenated file, you're guaranteed to have the newer node versions replace the older ones (this is exactly what's done in addnodes.sh). Sweet, huh? :-) Also included are a few other scripts I use to start/stop the node, and to create a new freenet.conf.defaults file that is totally unaffected by the current freenet.conf settings. Useful for later diff'ing to see what's new or changed. Here's a sample of an addnode.sh run: Backing up noderefs.txt Fetching noderefs.txt... Running dedupe.sh on noderefs.txt... Backing up noderefs.txt Collecting node version info... 128 node references found in noderefs.txt Deduping... Saved 128 unique nodes to noderefs.txt Backing up seednodes.ref Creating combined noderefs/seednodes file in seednodes.ref.new Running dedupe.sh on seednodes.ref.new... Backing up seednodes.ref.new Collecting node version info... 327 node references found in seednodes.ref.new Deduping... Saved 203 unique nodes to seednodes.ref.new Moving deduped seednodes.ref.new to seednodes.ref Done! Hope you'll find this stuff useful. I'm quite pleased with the results, myself. :-) Conrad -- Conrad Sabatier [EMAIL PROTECTED] - In Unix veritas scripts.tar.gz Description: scripts.tar.gz ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]
Re: [freenet-dev] Re: [freenet-support] Compilation error
On Wed, Feb 18, 2004 at 09:10:11AM -0800, Sheldon Young wrote: javac -target 1.1 -classpath lib/freenet-ext.jar:lib/junit.jar -sourcepath src -d build -deprecation src/freenet/client/*.java src/freenet/client/cli/*.java src/freenet/support/servlet/http/HttpSessionImpl.java:7: warning: javax.servlet.http.HttpSessionContext in javax.servlet.http has been deprecated import javax.servlet.http.HttpSessionContext; ^ 1 warning I don't know what it mean It means you should turn off -deprecation if you don't want to see the message :) Seriously though, this is annoying :| This is the last and only deprecation warning. Unfortunately it cannot be eliminated because, according to the Servlet 2.2 specification, getSession() must still exists in the HttpServletRequest class. As long as getSession() exists then the HttpSession class must exist, which means the deprecated class HttpSessionContext must exist. We do not however have to implement it. Just return null, and don't implement HttpSessionContext. Right? It's a little ugly, but it's harmless, although it may point out we are reinventing the wheel and maybe should be borrowing someone elses servlet implementation. -- Matthew J Toseland - [EMAIL PROTECTED] Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so. signature.asc Description: Digital signature ___ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]