LichtJF
06-08-2005, 12:09 AM
as found here: http://episteme.arstechnica.com/eve/ubb.x/a/tpc/f/122097561/m/817009733731/p/5
I know the question everyone is dying to ask...
"How is BuddhaMan able to get such high # of results from only three clients?"
Easy, I hax0red the xml config file for the client decreasing the traceroute time to one second from the slider's limitation of 15 seconds. There really is no difference in the result count when set somewhere between 1-5 seconds and it looks to my untrained eye that my three boxes are tracerouting as fast as they can.
How to up the trace frequency:
1) Make a backup of ..\DIMES\Agent\Classes\Base\conf\Properties.xml before even attempting this.
2) Open up ..\DIMES\Agent\Classes\Base\conf\Properties.xml with an editor (I used HTML-Kit myself).
3) Scroll down to the <scheduler> section which has two sub sections of <delay> and <period> which will probably have "15000" in there (if you set the slider to "15" already). The # is in thousandths, so set it to 1000 for one second, etc, etc.
4) Wait until the \DIMES\Agent\Classes\Base\Outgoing\results\xxxxxxxxxxxxxx.results.out.tmp file gets purged of it's info. You'll see this happen if you have the directory open and you see another file appear. This new file is caching the results received while the other data is sent back to DIMES HQ. Once you're back to only one file in that directory, you know your last set of results have been sent.
5) Restart the client. Besides reloading the properies.xml and using your new settings, this has the side affect of killing off your results file and starting it over. Hence the reason for step 4 above (I got yer back! Big Grin)
Notes: I have my 3 clients set to 5 seconds. Setting them any lower didn't seem to make the results come any faster. This may be due to the upstream bandwidth on my cablemodem only being 512 kbit. This brings me to my next point, bandwidth used.
I have a Linux firewall with IPTraf installed and accidentally left it (iptraf) running for way longer than I meant too...which turned out to be a smart thing because I now have data to crunch. Big Grin
So, running three clients with the DIMES clinet set to run a test every 5 seconds, my firewall detected the following ICMP traffic:
Outgoing: 14,647,814 bytes
Incoming: 19,918,610 bytes
Total: 34,589,681 bytes (#s were changing as I wrote the numbers down...we're not splitting atoms here)
Time: 25:39 (hrs/mins)
Math In Public results in: ~925 Mbytes of bandwidth used in 30 days (7.4Gbit) (I hope that's right...no flames for bad math Big Grin).
Additionally: there are other <delay> and <period> values in properties.xml config file. I tried tweeking the comState values, but saw no changes in the number of results received.
Improve upon my work and report back.
If I see someone rocket up the standings after today, I'll assume what happened. Big Grin
A Linux client would rock as I have a VPS server sitting on the Atlanta InterNAP with a fat pipe and 50GB bandwidth a month that's going almost unused at this point. :/
I know the question everyone is dying to ask...
"How is BuddhaMan able to get such high # of results from only three clients?"
Easy, I hax0red the xml config file for the client decreasing the traceroute time to one second from the slider's limitation of 15 seconds. There really is no difference in the result count when set somewhere between 1-5 seconds and it looks to my untrained eye that my three boxes are tracerouting as fast as they can.
How to up the trace frequency:
1) Make a backup of ..\DIMES\Agent\Classes\Base\conf\Properties.xml before even attempting this.
2) Open up ..\DIMES\Agent\Classes\Base\conf\Properties.xml with an editor (I used HTML-Kit myself).
3) Scroll down to the <scheduler> section which has two sub sections of <delay> and <period> which will probably have "15000" in there (if you set the slider to "15" already). The # is in thousandths, so set it to 1000 for one second, etc, etc.
4) Wait until the \DIMES\Agent\Classes\Base\Outgoing\results\xxxxxxxxxxxxxx.results.out.tmp file gets purged of it's info. You'll see this happen if you have the directory open and you see another file appear. This new file is caching the results received while the other data is sent back to DIMES HQ. Once you're back to only one file in that directory, you know your last set of results have been sent.
5) Restart the client. Besides reloading the properies.xml and using your new settings, this has the side affect of killing off your results file and starting it over. Hence the reason for step 4 above (I got yer back! Big Grin)
Notes: I have my 3 clients set to 5 seconds. Setting them any lower didn't seem to make the results come any faster. This may be due to the upstream bandwidth on my cablemodem only being 512 kbit. This brings me to my next point, bandwidth used.
I have a Linux firewall with IPTraf installed and accidentally left it (iptraf) running for way longer than I meant too...which turned out to be a smart thing because I now have data to crunch. Big Grin
So, running three clients with the DIMES clinet set to run a test every 5 seconds, my firewall detected the following ICMP traffic:
Outgoing: 14,647,814 bytes
Incoming: 19,918,610 bytes
Total: 34,589,681 bytes (#s were changing as I wrote the numbers down...we're not splitting atoms here)
Time: 25:39 (hrs/mins)
Math In Public results in: ~925 Mbytes of bandwidth used in 30 days (7.4Gbit) (I hope that's right...no flames for bad math Big Grin).
Additionally: there are other <delay> and <period> values in properties.xml config file. I tried tweeking the comState values, but saw no changes in the number of results received.
Improve upon my work and report back.
If I see someone rocket up the standings after today, I'll assume what happened. Big Grin
A Linux client would rock as I have a VPS server sitting on the Atlanta InterNAP with a fat pipe and 50GB bandwidth a month that's going almost unused at this point. :/