Sprints: Difference between revisions

From Nsnam
Jump to navigation Jump to search
Line 28: Line 28:
** '''Progress:''' The merge request is updated; failing tests need to be debugged
** '''Progress:''' The merge request is updated; failing tests need to be debugged
* [https://gitlab.com/nsnam/ns-3-dev/-/merge_requests/377 !377 Add FqPIE with an L4S mode]
* [https://gitlab.com/nsnam/ns-3-dev/-/merge_requests/377 !377 Add FqPIE with an L4S mode]
** '''Progress:''' The merge request is updated; all tests pass
* [https://gitlab.com/nsnam/ns-3-dev/-/merge_requests/373 !373 TCP Prague model]
* [https://gitlab.com/nsnam/ns-3-dev/-/merge_requests/373 !373 TCP Prague model]
** See below section on dual queue patch refresh
** See below section on dual queue patch refresh
Line 66: Line 67:
** '''progress:''' This branch is now at https://gitlab.com/tomhenderson/ns-3-dev/-/tree/dual-queue-coupled-aqm.  It runs the example, and the test suite runs (with some failures)
** '''progress:''' This branch is now at https://gitlab.com/tomhenderson/ns-3-dev/-/tree/dual-queue-coupled-aqm.  It runs the example, and the test suite runs (with some failures)
* test with the TCP Prague model developed by Deepak K in 2020 GSoC (https://gitlab.com/deepakkavoor/ns-3-dev/-/tree/prague-mr)
* test with the TCP Prague model developed by Deepak K in 2020 GSoC (https://gitlab.com/deepakkavoor/ns-3-dev/-/tree/prague-mr)
** '''progress:'''  I have merged this Prague model with the dual-queue-coupled-aqm branch to create a 'hackathon-ietf-110' branch
* test with the L4S evaluation suite (https://gitlab.com/tomhend/modules/l4s-evaluation)
* test with the L4S evaluation suite (https://gitlab.com/tomhend/modules/l4s-evaluation)
** '''progress:'''  The tcp-prague.cc example is based on this evaluation suite


The ultimate goal of all of the above is to have updated modules to experiment within the scenarios being used in the [https://datatracker.ietf.org/wg/tsvwg/about/ IETF tsvwg] to evaluate L4S, as described here:  https://l4s.cablelabs.com/
The ultimate goal of all of the above is to have updated modules to experiment within the scenarios being used in the [https://datatracker.ietf.org/wg/tsvwg/about/ IETF tsvwg] to evaluate L4S, as described here:  https://l4s.cablelabs.com/

Revision as of 16:51, 4 March 2021

Main Page - Roadmap - Summer Projects - Project Ideas - Developer FAQ - Tools - Related Projects

HOWTOs - Installation - Troubleshooting - User FAQ - Samples - Models - Education - Contributed Code - Papers

What is a sprint?

A sprint is a scheduled time for ns-3 contributors to gather and focus their undivided attention on improving the code, documentation, or tests.

Sprints are open to anyone, and have been organized on Zulip and IRC, and if needed, other virtual meeting technologies.

The Django project has a good reference and links to some presentations all about coding sprints.

Future sprints

IETF 110 Hackathon, March 1-4, 2020

  • Topic: L4S and traffic control models for ns-3
  • Time: 14:30-18:30 UTC, March 1-4 (Mon-Thurs.).
  • Chat: We will use the Hackathon stream on ns-3's Zulip channel.
  • Video: We are going to use Galene, a browser-based open source tool, courtesy of Juliusz Chroboczek. No password needed; just enter your name and Connect.
  • Participating: Tom Henderson, Mohit Tahiliani, others TBD
  • Contact: Tom Henderson <tomh at tomh.org>

We will focus on a few topics related to improved support of L4S and traffic controls in ns-3, along with experimentation with Linux on testbeds. L4S is an IETF research topic.

Several GSoC projects have pending merge requests, so adding them to the ns-3 mainline is a priority:

Some other related patches are in WIP form and could be updated if there is someone interested to work on them:

Testing and validation of Prague and L4S can be done with real implementations. There is a Linux kernel implementation of Prague but it is for kernel 5.4. Time permitting, we will also look at rebasing Prague to kernel 5.10 version for Google's BBRv2: https://github.com/google/bbr/tree/v2alpha

We will also look at freshening the CloudLab-based experimentation framework prevoiusly worked on here:

as well as a netns-based framework:

There are some bugs filed on our TCP model that could be looked at by anyone interested.

  • TCP issues. In particular, 59, 78, 125, 248.
    • Sachin is working on testing for issue #59

Dual queue patch refresh

Tom Henderson is working on refreshing the dual queue AQM for L4S. This model was originally written by Shravya KS in 2017 GSoC, but I have been working on it since then.

The ultimate goal of all of the above is to have updated modules to experiment within the scenarios being used in the IETF tsvwg to evaluate L4S, as described here: https://l4s.cablelabs.com/

Past sprints

We had several past sprints; details from the last few are listed below. Details from older ones can be found in this wiki page history.

IETF 109 Hackathon, Nov 9-12, 2020

  • Topic: TCP models for ns-3
  • Time: 15:00-19:00 UTC, Nov 9-12 (Mon-Thurs.).
  • Chat: We used the Hackathon stream on ns-3's Zulip channel.
  • Participating: Tom Henderson, Mohit Tahiliani, Aditya Chaudhary, Sayali Patil, Sushma Meena
  • Contact: Tom Henderson <tomh at tomh.org>

Summary slides: https://www.nsnam.org/wp-content/uploads/2020/ietf-109-hackathon-presentation-ns-3.pdf

TCP Cubic

Finish off the testing and validation work described here: https://gitlab.com/nsnam/ns-3-dev/-/merge_requests/408

This Google Document contains links to the latest status and steps to reproduce.

Accomplishments:

  • implement CWR state for ECN Echo handling
  • debug slow start window growth differences

BBRv1

Accomplishments: Further alignment against Linux BBRv1 results.

Flent

Harsha Sharma has been working on a flent application for ns-3 (to produce datasets similar to and compatible with the flent tool.

Accomplishments: Debug high latency of pings in RRUL test (due to hash collisions). Debug TCP throughput differences for RRUL tests (due in part to socket buffer size differences).

IETF 108 Hackathon, July 20-23, 2020

Our plan was to continue with testing, documenting, and preparing for the mainline, the various components that make up L4S support for ns-3

Integrated code was posted at the following two locations:

Some accomplishments for the week are listed below:

  • Deepak added RTT independence feature to the Prague code, and helped with some integration issues regarding AccECN and ECN++ code
  • Bhaskar's Fq/CoDel code was added to the integrated code above, and he further worked on a Fq/Cobalt variant
  • Harsha worked on Flent application support, exploring C++ JSON serialization libraries and how to use Flent to plot ns-3 output data
  • Ashutosh and Fraida created scripts and documentation for using CloudLab to test L4S kernel code
  • Tom performed overall ns-3 integration and testing. AccECN and ECN++ support had to be backed out due to some failing mainline tests, but existing Prague does not yet need the full AccECN support.

Below are some links for those who may want to obtain, review, and run the ns-3 code:

GSoC 2020 Code Sprint, May 16, 2020

  • Topic: ns-3 issues and merge requests
  • Time: 10:00 through 18:00 CEST (UTC+2)
  • Place: Online only. We will interact on ns-3 Zulip and on Jitsi.
  • Participating: Tommaso Pecorella, Ananthakrishnan Saji, Deepak K.

IETF 107 Hackathon, March 21-22, 2020

  • Unfortunately, this has been cancelled along with IETF 107.

IETF 106 Hackathon, November 16-17, 2019

  • Topic: IETF L4S models for ns-3
  • Time: 8:30am Singapore (UTC+8), Saturday Nov. 16, through 16:00pm Sunday Nov. 17
  • Place: The physical location was IETF 106 in Singapore. We will interact on ns-3 Zulip and on the L4S team's Slack channel.
  • Participating: Tom Henderson, Mohit Tahiliani, Vivek Jain, Joakim Misund

Our plan was to continue with testing, documenting, and preparing for the mainline, the various components that make up L4S:

A new ns-3 extension module for L4S experiments has been posted here.

Accomplishments

1) Update DualQ implementation to latest IETF draft.

The latest IETF draft is here: https://www.ietf.org/id/draft-ietf-tsvwg-aqm-dualq-coupled-10.txt

We want the ns-3 implementation to align as closely as possible to Appendix A of the draft, Figures 2-6 (i.e. without overload code for the first phase). This includes some variable naming updates, some arithmetic alignment, and reviewing the notes in Appendix A.

For example, the method recur(likelihood) was introduced in recent drafts.

Status: Tom Henderson completed this model port (still requiring tests and documentation update):

2) Create an initial TcpPrague model

Using the paced chirping code from Joakim Misund, take the current TcpDctcp model and create a TcpPrague model that initially is identical to TcpDctcp. Then, add paced chirping as a first step. This can be optionally enabled via an ns-3 attribute "EnablePacedChirping".

Status: Joakim continues to develop his branch: https://gitlab.com/JoakimMisund/l4s-evaluation/tree/tcp-prague

3) TCP ECN refactoring.

We are reworking the TCP ECN configuration code so that we can more flexibly configure ECT(0) or ECT(1) and ECN mode (classic or DCTCP or AccECN mode) and trigger it from the selection of congestion control

Status: After much discussion during the hackathon, Vivek Jain completed two patches that should be readied for the mainline:

IETF 105 Hackathon, July 20, 2019

  • Topic: IETF L4S models for ns-3
  • Participating: Ankit Deepak, Tom Henderson, Vivek Jain, Viyom Mittal, Mohit Tahiliani

A document describing the goals and plan was provided at this link.

Accomplishments

In the July 2019 IETF 105 hackathon sprint, we focused on updating a number of previously submitted L4S models.

Friday March 27, 2015, 13h00-19h00 UTC

 ./waf --run "third --PrintHelp"
  Waf: Entering directory `.../ns-3-allinone/ns-3-dev/build'
  Waf: Leaving directory `.../ns-3-allinone/ns-3-dev/build'
  'build' finished successfully (3.209s)
  ns3-dev-third-debug [Program Arguments] [General Arguments]

  Program Arguments:
    --nCsma:    Number of "extra" CSMA nodes/devices [3]
    --nWifi:    Number of wifi STA devices [3]
    --verbose:  Tell echo applications to log if true [true]

  General Arguments:
    --PrintGlobals:              Print the list of globals.
    --PrintGroups:               Print the list of groups.
    --PrintGroup=[group]:        Print all TypeIds of group.
    --PrintTypeIds:              Print all TypeIds.
    --PrintAttributes=[typeid]:  Print all attributes of typeid.
    --PrintHelp:                 Print this help message.

This command should also work; 

 ./waf --run "third --PrintGroups"
  Waf: Entering directory `.../ns-3-allinone/ns-3-dev/build'
  Waf: Leaving directory `.../ns-3-allinone/ns-3-dev/build'
  'build' finished successfully (3.231s)
  Registered TypeId groups:
    
    Mobility

We then want to be able to list all of the relevant TypeIds: 

./waf --run "third --PrintGroup=Mobility"
Waf: Entering directory `.../ns-3-allinone/ns-3-dev/build'
Waf: Leaving directory `.../ns-3-allinone/ns-3-dev/build'
'build' finished successfully (3.202s)
TypeIds in group Mobility:
    ns3::GaussMarkovMobilityModel
    ns3::GridPositionAllocator
    ns3::RandomBoxPositionAllocator
    ns3::RandomDirection2dMobilityModel
    ns3::RandomDiscPositionAllocator
    ns3::RandomRectanglePositionAllocator
    ns3::RandomWalk2dMobilityModel
    ns3::RandomWaypointMobilityModel
    ns3::SteadyStateRandomWaypointMobilityModel
    ns3::UniformDiscPositionAllocator
    ns3::WaypointMobilityModel

Then the attributes of a given class; this type of capability will be enabled across the codebase if we set the Group names consistently: 

 ./waf --run "third --PrintAttributes=ns3::RandomWalk2dMobilityModel"
Waf: Entering directory `.../ns-3-allinone/ns-3-dev/build'
Waf: Leaving directory `.../ns-3-allinone/ns-3-dev/build'
'build' finished successfully (3.187s)
Attributes for TypeId ns3::RandomWalk2dMobilityModel
    --ns3::RandomWalk2dMobilityModel::Bounds=[0|100|0|100]
        Bounds of the area to cruise.
    --ns3::RandomWalk2dMobilityModel::Direction=[ns3::UniformRandomVariable[Min=0.0|Max=6.283184]]
        A random variable used to pick the direction (radians).
    --ns3::RandomWalk2dMobilityModel::Distance=[1]
        Change current direction and speed after moving for this distance.
    --ns3::RandomWalk2dMobilityModel::Mode=[Distance]
        The mode indicates the condition used to change the current speed and direction
    --ns3::RandomWalk2dMobilityModel::Speed=[ns3::UniformRandomVariable[Min=2.0|Max=4.0]]
        A random variable used to pick the speed (m/s).
    --ns3::RandomWalk2dMobilityModel::Time=[+1000000000.0ns]
        Change current direction and speed after moving for this delay.
  • Tasks: We will organize to group as many TypeIds as possible, across module lines. For instance, someone will be assigned the src/applications module, someone will be assigned the src/propagation model, and so on. When finished or having taken it as far as you can, submit a patch for the module.
  • Patches: We would like code to be sent in unified diff format. An easy way to do this is to use the 'hg diff' command, and redirect the output into a file (and submit the file). For instance, here is the output of 'hg diff' if one line of text is added:
hg diff
diff -r 6ea11eb86c17 src/applications/model/application-packet-probe.cc
--- a/src/applications/model/application-packet-probe.cc	Thu Mar 19 21:13:16 2015 +0100
+++ b/src/applications/model/application-packet-probe.cc	Mon Mar 23 15:44:33 2015 -0700
@@ -39,6 +39,7 @@
 {
   static TypeId tid = TypeId ("ns3::ApplicationPacketProbe")
     .SetParent<Probe> ()
+    .SetGroupName ("Application")
     .AddConstructor<ApplicationPacketProbe> ()
     .AddTraceSource ( "Output",
                       "The packet plus its socket address that serve "

If you type 

 hg diff > application.patch

you will store the above in a patch file, which can be submitted.

Please remember that an implementation file (.cc) may contain multiple TypeId declarations, so please search through each file.

  • Outcome: If successful, the PrintGroups command will have many groups defined (one for each module, perhaps more). The tutorial and manual will also be extended to describe how to use this feature.
  • Coordination: We will use IRC and Etherpad to coordinate the work (make sure multiple people do not work on the same file).

Accomplishments

In the March 2015 sprint, we were able to commit about half of the SetGroupName() methods across the codebase. Following the sprint, the remainder of the codebase was handled, so by the ns-3.23 release, this item was completed

Friday December 12, 2014, 14h00-20h00 UTC

Fix as many of the doxygen warnings as we can. A current count of warnings (by module) can be seen here: http://mailman.isi.edu/pipermail/ns-commits/2014-November/015994.html. There is a full log generated that can be used to identify what is the error, and then it can be fixed with a text editor. How this is planned to work is that contributors who show up will choose a file or set of files (that they are comfortable with documenting) and will try to clear the errors in that file, producing a patch that can be merged to ns-3-dev.
We may also work on reducing includes in ns-3.
  • Coordination: We will use IRC and Etherpad to coordinate the work (make sure multiple people do not work on the same file).

Accomplishments

In the December 2014 sprint, we were able to get to zero warnings on the LTE helper directory, and also the mobility model and helper directories. December sprint warnings fixed (Peter, Piotr, Tom):

  • 176 src/mobility/model (all eliminated)
  • 39 src/mobility/helper (all eliminated)
  • 146 src/lte/helper (all eliminated)
  • reduced src/core warnings to below 500 count

October 2014 Doxygen sprint

This was a Doxygen-focused sprint with the following instructions.

  • Use mercurial to check out the current copy of ns-3 (see the tutorial if you don't know how to do this):
  • You must also have Doxygen installed on your machine.

Once you have an ns-3-dev/ directory, take these steps: 

   $ ./waf configure --enable-examples --enable-tests
   $ ./waf build

Then, run the script doc/doxygen.warnings.report.sh from the ns-3-dev directory: 

 $ ./doc/doxygen.warnings.report.sh -t -e

(Note: this excludes tests and examples; to include warnings for those files, rerun without the -t and -e options). This will produce lots of summary statistics like: 

  Warnings by module/directory:

  Count Directory
  ----- ----------------------------------
  3390 src/lte/model
  1532 src/wimax/model
   575 src/core/model
   570 src/wifi/model
   488 src/mesh/model
   338 src/visualizer/visualizer
   229 src/netanim/model
   213 src/dsr/model
   ...

   ...

Do not worry if the output scrolls beyond the buffer in your terminal window; the full warnings log is written to doc/doxygen.warnings.log.

Once this is complete, you can use the doxygen.warnings.report.sh script to focus on one module or file. To see just the errors from one module, for example wifi: 

 $ ./doc/doxygen.warnings.report.sh -t -e -s -m wifi

The -s option skips the doxygen run, and just reports results from the existing warnings log. Obviously, if you want to see the results of your edits, you would omit the -s option so you could regenerate the warnings log. The -m <module> option only shows results from the ./src/<module> module; after printing the warnings counts by file, it shows the actual errors from each file: 

 ...
 Filtered Warnings
 ========================================
 src/wifi/helper/athstats-helper.h:49: warning: Member EnableAthstats(std::string filename, uint32_t nodeid, 
 uint32_t deviceid) (function) of class ns3::AthstatsHelper is not documented.
 ...

This means the method AthstatsHelper::EnableAthstats has no documentation, and it's declared in src/wifi/helper/athstats-helper.h around line 49.

To see just the warnings from a single file (or a set of files matching a regular expression) use the -F <file-name-or-regex> option: 

 $ ./doc/doxygen.warnings.report.sh -s -F "block-ack.*"


Once you find the list of files and look at the possible errors to fix, feel free to join the chat and request to take a module or file and fix it, and we will list it on the Etherpad as being taken. While all docs are valuable, we should probably focus on code in the model/ directories, then helpers/, examples/, and finally test/.

To add doxygen documentation, please follow the ns-3 Specifics guidelines.

Don't be surprised if you fix one warning and more appear! If a struct, class or function has no documentation, doxygen reports a single "is not documented" error. Once that object has documentation, then doxygen will examine it in detail and report additional warnings. For example, a function with no documentation will generate one warning. Adding documentation about the function itself will fix that warning, but then generate warnings for each of the function parameters and return value, until those are also documented.

Unfortunately, it's not enough just write documentation; you also have to test it. (Documentation has bugs too!) Just rerun the doxygen.warnings.report.sh script without the -s option: 

 $ ./doc/doxygen.warnings.report.sh -m <module> # or -F <file-name-or-regex>

(BTW, without -s the script will always build all the doxygen; the -m and -F options only affect what it displays from the resulting doxygen.warnings.log)

Once the script completes, load up the docs in your browser and check that it looks as you intended/you didn't mistype some doxygen syntax: 

 $ firefox doc/html/index.html &

As in all code development: wash, rinse, repeat, or, in this case: 

 $ ./doc/doxygen.warnings.report.sh -m <module>
 [edit, edit, edit]
 $ firefox doc/html/index.html &


Once you have fixed a file, you may create a patch such as follows: 

 $ hg diff file-you-have-fixed.h > file-you-have-fixed.h.patch

and upload it to bug 938 or email it to one of the maintainers participating in the sprint (you can ask on IRC about it).


  • Make sure that you can join the IRC channel in advance
  • Make sure you can checkout the development version of ns-3 (we will use ns-3-dev)
  • Familiarize yourself with how to generate a patch against ns-3-dev. Patches you write can be uploaded to Bugzilla or sent to one of the maintainers.

Accomplishments

In the October 2014 sprint, the following warnings were fixed (Budi, Tiago, Tommaso, Natale, Peter):

  • src/lte warnings count reduced by ~1500
  • src/mobility warnings count reduced by 20
  • all 160 propagation warnings eliminated
  • all 47 point-to-point warnings eliminated
  • all 55 internet warnings eliminated
Retrieved from "https://www.nsnam.org/mediawiki/index.php?title=Sprints&oldid=12213"