[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: [ns] NS with GPRS-module - successfully passed GPRS-module internal test (example.tcl)
--
Free Dmitry Sklyarov !
http://www.freesklyarov.org
> -----Original Message-----
> From: Robert Iakobashvili [SMTP:[email protected]]
> Sent: Sunday, October 21, 2001 12:45 PM
> To: [email protected]
> Cc: [email protected]; [email protected]; [email protected]
> Subject: [ns] NS with GPRS-module - successfully passed GPRS-module internal test (example.tcl)
> Importance: Low
>
> Yes, indeed.
>
> I have prepared a distribution file, named ns-allinone-2.1b7a.tar.gz,
> size about 49 M with
> a README_FROM_ROBERT file.
>
> You are welcome to tell me your anonymous ftp-site with write
> permissions, or any other account on your choice
> to make put to it.
>
> Sorry, but internal security arrangements do not enable me to offer you
> a web site or ftp site for download.
>
> What was done is :
> My OS is Solaris 5.7 on sparc, NS simulator version is
> ns-allinone-2.1b7a.
>
> I tried to follow the procedure of Roger ([email protected]) and did the
> following:
>
> STEP 1:
>
> Dowloaded the all-in-one-installation of ns based on ns-allinone-2.1b7a
> and already incorporating the NOAH
> extension from Jorg Widmer site
> http://www.isci.berkeley.edu/~widmer/mnav/ns-extension, named
> ns-wireless-2.1b7a.tar.gz .
>
>
> Another option is to copy manually files with Thomas Schwabe version of
> the NOAH extension from the same site into standard
> ns-wireless-2.1b7a.tar.gz distribution.
>
>
> KNOWN PROBLEMS (STILL PEOPLE LIVE WTH THAT, see posted by Roger's
> correpondence)
> In both cases the compilation was o'key, but ./validate failed for the
> tests:
>
> mip-adv-one, mip-adv-zero, mip-adv-multi failed with core dump;
> test-all-energy.tcl : dsdv, dsr, aodv - compute-routes problem ;
> dsdv-wired-cum-wireless has a meaningfull, but different output ;
> wireless1 and wireless2- configuration problems;
> wireless3 - core dump ;
> test-all-wireless-tdma.tcl : dsdv and dsr - configuration problem
> test-suite-wireless-shadowing.tcl : dsr - configuration and routing
> problem
> test-suite-wireless-diffusion.tcl : diff-rate-default, diff-rate-other,
> omnicast, flooding - configuration problems
>
>
> STEP 2. GPRS MODULE from Richa Jain
>
> To the directories obtained from step 1 :
> 1. applied patch pgrs_patch ;
> 2. copied files from final_code directory ;
> 3. compiled successfully ./install once more;
> 4. setting of all environment var ;
>
> Now the problem:
>
> 'ns example.tcl' test from GPRS module has failed with:
> rlc-timers.cc:88: failed assertion 'busy_' ;
>
> I commented the assert temporarily as well as assert in ll-timers.cc:87
>
> Modified output file instead of "gprs_err_acked.tr" to be
> "gprs_err_acked.tr.21.10.2001"
>
> and tryed 'ns example.tcl'
>
> cd to the directory with the GPRS package; cd gprs_doc
>
> gzip -d gprs_err_acked.tr.gz
>
> diff gprs_err_acked.tr FULL_PATH/gprs_err_acked.tr.21.10.2001
>
> NO DIFFERENCES REPORTED
>
>
> The build was copied to directory ns-allinone-2.1b7a-GPRS
>
> cded to directory and make uninstall ; make realclean or distclean or
> clean or veryclean
>
> This file was added to the ns-allinone-2.1b7a-GPRS directory.
>
> Please, pay attention to the above mentioned asserts, as I had no time
> to go inside the matter.
>
This is correct, I had to do the same. I wonder how it could ever work.
I also had once a more or less working version of the gprs module lying around, but I think I hacked it to death ;-)
Be warned, there are still problems with the gprs implementation:
the bitrate of a link using 1 time slot is 1/0.000577 s/slot /8 slots/frame *8 bit/byte *50 byte/slot = 86.6 kbit/s, for all 8 channels together this gives 693 kbps
I think this is a little bit much. If you change it from 50 bytes per slot to 20 bytes per slot, it should be better
fragmentation in the rlc layer: let's say you have a 120 byte packet. then the rlc layer will generate three fragments, 50 bytes, 50 bytes, 20 bytes, the remaining 30 bytes in the last packet stay empty. I don't think this is really good.
Let's say you have two users, #1 uses time slot 1, #2 uses timeslot 2.
Now you get 10 fragments from user #1. To send them you need 10 tdma frames. While sending these 10 fragments, no other timeslots can be used, so user #2 has to wait until #1 has finished, since there is only 1 queue between rlc and mac-gprs, so the 10 fragments of #1 have to be processed before other packets can be processed.
before a transmission starts, no matter whether voice or data, arp packets are exchanged, is this correct ???
Bye
Alex