Download Network Simulator (NS) 2.34 and unzip it.
$ tar -zxvf ns-allinone-2.34.tar.gz
Copy pFabric.patch to the top ns-2.34 folder (ns-allinone-2.34) and apply the patch. Then install NS2.
$ cd ns-allinone-2.34
$ patch -p1 --ignore-whitespace -i pFabric.patch
$ ./install
Copy files in tcp folder to ns-allinone-2.34/ns-2.34/tcp/.
Copy files in queue folder to ns-allinone-2.34/ns-2.34/queue/.
Add queue/priority.o to ns-allinone-2.34/ns-2.34/Makefile.
Run make on /ns-allinone-2.34/ns-2.34.
Please see our latest online version which is modified from the original NSDI version. In this new version, we update demotion threshold derivation and fix some bugs that may degrade the performance of PIAS.
You can find simulation scrips in scripts folder:
- CDF_dctcp.tcl (web search workload), CDF_vl2.tcl (data mining workload) give flow size distributions used in our paper.
- spine_empirical.tcl, spine_empirical_heter.tcl (heterogeneous traffic pattern) and [tcp-common-opt.tcl] (https://github.com/HKUST-SING/PIAS-NS2/blob/master/scripts/tcp-common-opt.tcl) are NS2 TCL simulation scripts.
- result.py is used to parse final results.
- run_pias_dctcp.py is used to run PIAS and DCTCP (one-priority PIAS) using the web search workload.
- run_pias_vl2.py is used to run PIAS and DCTCP (one-priority PIAS) using the data mining workload.
- run_pfabric_dctcp.py is used to run pFabric (remaining size and bytes sent) using the web search workload.
- run_pfabric_vl2.py is used to run pFabric (remaining size and bytes sent) using the data mining workload.
- run_lldct_dctcp.py is used to run L2DCT using the web search workload.
- run_lldct_vl2.py is used to run L2DCT using the data mining workload.
- run_pias_dctcp_oversub.py is used to run PIAS and DCTCP (one-priority PIAS) using the web search workload in a 3:1 oversubscribed network.
- run_pias_heter.py is used to run PIAS and DCTCP (one-priority PIAS) using the heterogeneous workload (a mix of web search and data mining).
There are many parameters to configue in run_[transport]_[workload].py. Note that you need to modify ns_path and sim_script correspondingly.
For each simulation, it will create a folder whose name is [workload]_[transport]_[load]. For example, if we run PIAS using the websearch workload at 90% load, we will get a folder named websearch_pias_90.
Each folder contains two files: flow.tr and logFile.tr. The flow.tr gives flow completion time results with the following format:
number of packets, flow completion time, number of timeouts, src ID, dst ID
You can use result.py to parse flow.tr files as follows:
$ python result.py -a -i [path]/flow.tr
If you have any question about PIAS simulation code, please contact Wei Bai.
We thank Mohammad Alizadeh for sharing pFabric simulation code.