Repository navigation
showcases: add: a 10BASE-T1S showcase that compares PLCA with CSMA/CD - #1468
Conversation
INET models the 10BASE-T1S multidrop segment of IEEE 802.3cg with Physical Layer Collision Avoidance (PLCA), and examples/ethernet/TenBaseT1S holds its test and analysis configurations, among them a comparison with CSMA/CD in validation.ini. The user's guide chapter on 10BASE-T1S describes the modules and their statistics, but no page shows what PLCA changes on a shared bus, so a reader cannot learn from INET why PLCA is used in automotive and industrial networks, or how to set up and measure such a segment. The new showcase showcases/ethernet/tenbaset1s, "Bounded Channel Access on 10BASE-T1S: PLCA versus CSMA/CD", runs the same bus, nodes and traffic with PLCA and with CSMA/CD, switched by one configuration line. The page follows three properties of PLCA: a bounded wait for the channel, no frame lost to arbitration, and equal shares at saturation. With 8 nodes and 64-byte frames, the worst observed access latency of PLCA stays under the 614 us cycle bound at every load, while that of CSMA/CD grows to about 0.2-0.3 s and CSMA/CD drops 3-33 frames per run at the retry limit at 70 % and 85 % load. It uses the retry limit of 16 attempts and the retry-limit drop statistic that #1311 added. The showcase has seven configurations, an analysis file with seven charts, and a page in the Ethernet category, which the showcases index lists under the latest updates. Its images are in the media repository, and doc/media links to them. Baseline changes: fingerprint runs added (7; showcases/ethernet 7) — the commit adds the showcase and its seven configurations The statistics and chart pins move to baseline commits that only add the files of the new showcase, with the form first-record: statistics 63196dfc0b and media 83ffedd1e8, on the branch topic/gy/tenbaset1s-showcase of each repository. Change: examples.showcases | behavior.add | fingerprint statistical chart
There was a problem hiding this comment.
Devin Review found 2 potential issues.
1 flag not posted on this PR by your GitHub settings — view it in Devin Review. (Configure)
| sim-time-limit = 2ms | ||
| **.vector-recording = true | ||
| *.numNodes = 8 | ||
| *.controller.numApps = 9 |
There was a problem hiding this comment.
🟡 Repeated controller application counts
numApps repeats the network size in five configurations instead of deriving the required sink-plus-sources count. This violates AR-CFG-INFER, which requires deriving structural configuration rather than repeating it.
Learn more
The controller needs one receiving application and one sending application per edge node. CycleAnatomyBase, CsmaCdLatency, Scale40Plca, PlcaUtilization, and CsmaCdUtilization each encode this relationship with separate numeric literals. PlcaLatency instead repeats its iteration variable in the application count. These values currently agree, but they duplicate a structural fact that AR-CFG-INFER requires deriving.
Example: CycleAnatomyBase sets eight edge nodes and nine controller applications separately; the latter is entirely determined by the former.
Recommended fix: Define the controller application count once in General as an expression derived from the enclosing network's numNodes parameter, and remove the per-configuration assignments. Verify that the seven runnable configurations retain their existing effective application counts and fingerprints.
Was this helpful? React with 👍 or 👎 to provide feedback.
| [Config CycleAnatomy] | ||
| extends = CycleAnatomyBase, PlcaBase | ||
|
|
||
| [Config CsmaCdCycleAnatomy] | ||
| extends = CycleAnatomyBase, CsmaCdBase |
There was a problem hiding this comment.
🟡 Missing descriptions for every configuration
All ten named configurations omit description, with none inherited from General. This violates NR-INI, which requires a description for every configuration.
Learn more
The new file defines three abstract configurations and seven runnable configurations. None defines a description, and General does not provide one. NR-INI requires every configuration to carry a description.
Example: CycleAnatomy and CsmaCdCycleAnatomy expose only their configuration names rather than descriptions distinguishing their arbitration schemes.
Recommended fix: Add a concise description to each named configuration, identifying its arbitration scheme, experiment purpose, and relevant sweep or traffic setting. Fix these omissions in the configurations rather than adding an exception-ledger entry.
Was this helpful? React with 👍 or 👎 to provide feedback.
INET models the 10BASE-T1S multidrop segment of IEEE 802.3cg with Physical Layer Collision Avoidance (PLCA), but no page shows what PLCA changes on a shared bus. The user's guide chapter describes the modules and their statistics, and examples/ethernet/TenBaseT1S holds test and analysis configurations, so a reader cannot learn from INET why PLCA is used in automotive and industrial networks, or how to set up and measure such a segment.
This pull request adds the showcase "Bounded Channel Access on 10BASE-T1S: PLCA versus CSMA/CD" (showcases/ethernet/tenbaset1s). It runs one bus, one set of nodes and one traffic load with PLCA and with CSMA/CD, switched by a single configuration line, and follows three properties of PLCA: a bounded wait for the channel, no frame lost to arbitration, and equal shares at saturation. With 8 nodes and 64-byte frames, the worst observed access latency of PLCA stays under the 614 us cycle bound at every load, while CSMA/CD reaches about 0.2-0.3 s and drops 3-33 frames per run at the retry limit at 70 % and 85 % load. It relies on the 16-attempt retry limit and the retry-limit drop statistic of #1311, which are on master.
Risk: none for the model. The change adds files under showcases/, two index entries, seven fingerprint runs and the two baseline pins; no source changes.
Commits
Architectural surface: none. No contract, packet, parameter or feature descriptor changes; the showcase uses existing modules (EthernetPlcaHost, the EthernetPlca sublayer, EthernetCsmaMac).
Evidence
All in release mode, on the branch head, with the opp_repl runners (
--no-build, working-directory filtershowcases/ethernet/tenbaset1s):opp_update_fingerprint_test_results --no-stored-only -r '^0$' --exclude-config-filter '^General$'inserted 7 runs (tplx, ~tNl, ~tND);opp_run_fingerprint_teststhen gives 7 PASS.opp_update_statistical_test_results --exclude-config-filter '^General$'inserted 7 files;opp_run_statistical_teststhen gives 7 PASS.opp_update_chart_test_resultsdrew the 7 charts of TenBaseT1S.anf;opp_run_chart_teststhen gives 7 PASS.doc/project/enforcement/check-pull-request.sh --base origin/masterandcheck-baselines.sh origin/master..HEAD --baseline-repository statistical=statistics --baseline-repository chart=media: pass.doc/src/build --only showcases: no warning for the new page; all figures and downloads resolve.The candidate values were recorded on a local machine, not in the CI environment. If the statistical or chart tests of this pull request fail on these files, the CI artifacts statistical-candidates and chart-candidates replace them in the baseline commits. The General config of the showcase is not complete on its own (numNodes is unset), so it gets no baseline and the statistical test reports it as an error. The fingerprint debug leg was not run locally; Scale40Plca, the longest run, takes 66 s in release.
Baseline branches, for review of the values:
Baseline changes
7 fingerprint runs, 7 statistical tests and 7 chart images are added, all first records of the new showcase; no existing value moves.