Rust''s debug profile costs the bench ~45% of solo throughput and ~3x of
p99 latency vs release (267 vs 184 solo calls/sec, p99 5.7 vs 16ms).
Debug disables inlining, runs overflow checks on every arithmetic op,
keeps Future state machines un-collapsed, and lets every Vec allocation
through unoptimized. Other compiled clients in the matrix don''t see
this gap: Go always builds optimized, Python is interpreted, and the
JIT-tiered runtimes (HotSpot for Java, CoreCLR Tier 1 for .NET) close
most of the gap during the warmup window.
The driver now requests `cargo run --release` for Rust and `dotnet run
-c Release --no-build` for .NET, so the two compiled-AOT clients race
under their production-equivalent profiles. Callers must `cargo build
--release -p mxgw-cli` and `dotnet build ... -c Release` once before
running the bench; `--no-build` then keeps each measurement window
free of compilation overhead.
Live re-run (5-way concurrent, 30s, bulkSize 6) after the switch:
rust: 145.35 calls/sec (was 123.26 in debug; 18% gain under contention)
go: 185.59 calls/sec
java: 171.80 calls/sec
dotnet:172.31 calls/sec
python:140.52 calls/sec
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a bench-read-bulk subcommand to every client CLI (.NET, Go, Rust,
Python, Java) and a PowerShell driver that runs all five concurrently
against the deployed gateway and prints a side-by-side comparison.
Each CLI''s bench:
- Opens its own session, registers, subscribes to bulk-size tags so the
worker''s MxAccessValueCache populates from real OnDataChange events.
- Runs a warmup-seconds-long pre-loop with identical calls so JIT /
connection-pool / first-call overhead is amortised before the
measurement window.
- Runs ReadBulk in a tight in-process loop for duration-seconds with
per-call high-resolution latency capture (Stopwatch in .NET,
time.Now in Go, std::time::Instant in Rust, time.perf_counter in
Python, System.nanoTime in Java).
- Unsubscribes + closes the session, then emits one JSON object with
the shared schema: { language, durationMs, totalCalls, successfulCalls,
failedCalls, totalReadResults, cachedReadResults, callsPerSecond,
latencyMs: { p50, p95, p99, max, mean } }.
The PS driver (scripts/bench-read-bulk.ps1) launches one detached process
per client, waits for all to finish, parses the trailing JSON object from
each stdout, prints a comparison table, and persists the combined report
under artifacts/bench/. Quoting around Java''s `gradle --args="..."` is
handled by writing a one-shot .bat that cmd.exe runs; the .NET CLI''s
per-call gRPC timeout is auto-scaled to (Duration + Warmup + 30s) so the
channel-wide timeout doesn''t cancel the bench mid-loop.
Live 30-second steady-state run against the deployed gateway, all five
clients hitting the same six TestMachine_001..006.TestChangingInt tags:
client calls/sec cached/total p50 ms p95 ms p99 ms max ms
dotnet 171.78 30924/30924 3.84 14.06 40.41 542.48
go 175.46 31590/31590 3.93 13.52 41.26 243.00
rust 123.26 22188/22188 5.52 15.78 48.11 544.41
python 145.79 26244/26244 4.86 14.85 41.65 645.84
java 181.12 32604/32604 3.80 10.59 33.37 344.27
143,550 ReadBulk results across all five clients during the 30s window;
100% were was_cached = true (the worker''s cache fast-path never fell
through to the snapshot lifecycle). Aggregate read throughput ~800
calls/sec against five concurrent sessions sharing the same cached tags.
A second variant with bulk-size 20 sustained the same per-client call
rate while delivering 3.3x more values per call (~37,000 cached reads/sec
aggregate across the five concurrent sessions), confirming the linear
per-tag cache lookup inside one call is not a bottleneck at this scale.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The previous commit added the bulk read/write library surface in every
client; this commit makes that surface reachable from each client's CLI
and exercises it through scripts/run-client-e2e-tests.ps1.
Five new subcommands in every client CLI (.NET / Go / Rust / Python /
Java): read-bulk, write-bulk, write2-bulk, write-secured-bulk, and
write-secured2-bulk. Each follows the existing subscribe-bulk shape:
- read-bulk takes --server-handle, --items <csv tag list>, and
--timeout-ms (0 = worker default). JSON output carries the
BulkReadResult fields, including was_cached so the e2e matrix can
verify the cached-path semantics.
- The four bulk-write families take --server-handle, --item-handles
<csv>, --type, --values <csv>. write2-bulk and write-secured2-bulk
add a single --timestamp applied to every entry; the secured
variants take --current-user-id and --verifier-user-id. All four
output BulkWriteResult JSON.
A new -SkipReadWriteBulk switch on the matrix script (default OFF)
controls two new e2e phases:
- After the existing subscribe-bulk phase leaves tags advised, the
script runs read-bulk against the same tag list and asserts most
results return was_cached = true. This is the only e2e coverage of
the cache-then-snapshot fork — the unit + gateway tests verify the
semantics with a fake worker, but only the live cross-language
matrix proves the cache populates from real OnDataChange events and
survives the round-trip through every client''s JSON parser.
- When -VerifyWrite is set, the write phase now also runs a single-
entry write-bulk against the same writable item handle (using a
distinct sentinel value) and asserts a per-entry success. Confirms
the BulkWriteResult wire format end-to-end without complicating
the OnWriteComplete echo assertion the single-item phase already
verifies.
Dry-run validation passes for all five clients: each emits the correct
read-bulk and write-bulk CLI invocations with the right flags.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Running the matrix against a live gateway surfaced several issues:
- The write phase is now opt-in (-VerifyWrite, was -SkipWrite). It runs
right after register so only a small event backlog precedes the write,
and asserts the reliable OnWriteComplete signal (the written value is
not echoed back by a provider-driven attribute like TestChangingInt, so
the value compare is best-effort).
- Java was launched as bare "gradle", which .NET's Process.Start cannot
exec (it is gradle.bat) — resolve the launcher and run it via cmd.exe.
- The Java client's MxEventStream queue capacity was 16, which overflows
on any active session's backlog-replay burst; raised to 1024.
- The Rust stream-events CLI now renders the event family as the proto
enum name, matching the protobuf-JSON the other four clients emit.
Update docs/GatewayTesting.md for the reworked write phase.
Verified live: the full five-client matrix passes with -VerifyWrite.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Close the notable gaps in scripts/run-client-e2e-tests.ps1:
- Write round-trip: write a per-client sentinel value to a configurable
writable attribute, then assert it is echoed back through the event
stream. Extends the Rust mxgw-cli stream-events output with full
per-event JSON (itemHandle + protojson-shaped value) so all five
language clients run an identical value compare.
- Parity: assert an invalid item handle and an unknown session id are
rejected rather than silently succeeding.
- Auth rejection: assert open-session is rejected with a missing API key
and, when -RejectScopeApiKeyEnv is supplied, with an insufficient-scope
key.
- Parallel: -Parallel runs each language client as an isolated child
process and merges their JSON reports.
Update docs/GatewayTesting.md for the new phases and flags.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- regen-readme.py: use `python` not the broken `python3` Store alias in
the generated note and docstring; --check now also fails when a module
header's "Open findings" count disagrees with finding statuses or a
finding has an unrecognised Status (find_inconsistencies)
- REVIEW-PROCESS.md: rewritten for mxaccessgw (was describing ScadaLink)
— MxGateway.* modules, "mxaccessgw conventions" checklist category,
gateway.md/docs/ design context, `python` command
- scripts/check-code-reviews-readme.ps1: CI/pre-commit wrapper for
regen-readme.py --check
- code-reviews/test_regen_readme.py: dependency-free parser tests
- code-reviews/README.md: regenerated
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Rename 16 kebab-case docs to PascalCase per StyleGuide
- Move per-language client design docs from docs/ to clients/<lang>/
alongside their READMEs
- Add ## Related Documentation sections to 15 docs that lacked one
- Fix sentence-case violations in H3 headings (StyleGuide rule)
- Update cross-references in gateway.md, client READMEs, scripts,
and generate-proto.ps1 helpers to follow the new paths
- Add CLAUDE.md with build/test commands, the source-update
verification matrix, the parity-first contract, and pointers
to MXAccess and Galaxy Repository analysis sources
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Joseph Doherty