Centralize the bulk-transfer sizing in config.rs and bump the values used
on both ends of a QUIC connection:
- CHUNK_SIZE: 32 MiB -> 128 MiB
- QUIC_CONNECTION_DATA_WINDOW: 64 MiB -> 256 MiB
- QUIC_STREAM_DATA_WINDOW: 32 MiB -> 128 MiB
- QUIC_MAX_SEND_BUFFER_SIZE: 32 MiB -> 128 MiB
- QUIC_INITIAL_CONGESTION_WINDOW: 1 MiB -> 4 MiB
- FILE_TRANSFER_BUFFER_SIZE: 64 KiB -> 1 MiB (new constant)
The previous 32 MiB stream window was already comfortably above the
bandwidth-delay product of a sub-millisecond LAN at 2.5 GbE. The further
bump is deliberately generous: the goal is to push flow control and
per-syscall overhead far enough out of the way that they cannot be the
suspect when isolating the remaining LAN download bottleneck (disk, NIC,
or s2n-quic platform offload on the sending host). Memory pressure from
the larger windows is not observable on a desktop client moving GB-sized
games.
stream_file_bytes previously read the local file in 64 KiB chunks. At
multi-Gbit/s send rates that produced many thousands of disk reads per
second; bumping to 1 MiB keeps the per-file syscall load modest with no
measurable latency cost on streamed bulk transfers. The buffer size lives
in config.rs as FILE_TRANSFER_BUFFER_SIZE so it stays adjustable from one
place.
Also add a started/MiB-per-second log line at info level when a file
finishes streaming. This matches the S37 measurement methodology already
used in the peer-cli harness and makes per-file send throughput visible in
normal operation.
The peer-cli extended-scenarios harness uses CHUNK_SIZE as the tolerance
bound for chunk-boundary variance in its assertions, so its constant is
bumped to match. The multi-chunk planning unit test is rewritten to
reference CHUNK_SIZE symbolically (CHUNK_SIZE * 3 + CHUNK_SIZE / 2)
instead of a hardcoded 120 MiB; the previous literal would silently
degrade into a single-chunk test at the new chunk size and stop
exercising the spread-across-peers code path.
Test Plan:
- just fmt
- just clippy
- just test
- python3 crates/lanspread-peer-cli/scripts/run_extended_scenarios.py S37 \
--build-image
- python3 crates/lanspread-peer-cli/scripts/run_extended_scenarios.py S37
Refs: local LAN download performance investigation on 2026-05-20.
Depends-on: d7f7dc737e (QUIC UDP socket buffer sizing).
Add peer-cli accounting for download sessions so terminal download events
report bytes, chunks, elapsed time, MiB/s, and Mbit/s. The extended scenario
runner now has S37, a focused single-source download benchmark that creates a
2 GiB sparse bf1942 archive, downloads it from one peer with install disabled,
and checks the destination archive size and reported byte count.
This gives the QUIC performance work a repeatable measurement below the 5 GiB
limit from the original request. The source file is sparse, so S37 is aimed at
the app, QUIC, and destination-write path rather than raw source-disk reads;
the existing correctness scenarios still cover normal game downloads.
Baseline S37 before QUIC tuning:
- 733.22 MiB/s
- 6150.72 Mbit/s
- 2.793s for 2.00 GiB plus version.ini
- 65 reported chunks
Test Plan:
- just fmt
- python3 crates/lanspread-peer-cli/scripts/run_extended_scenarios.py S37 --build-image
Refs: local LAN download performance investigation on 2026-05-20.
Merge the S18-S36 scenario ideas into the official peer-cli scenario matrix
and add a Docker-backed runner that now exercises S1-S36 with concrete file
proofs. The runner creates temporary fixtures under .lanspread-peer-cli, drives
JSONL peer containers, checks transferred roots with diff and SHA-256 manifests,
and covers startup, discovery, transfer, failure, mutation, concurrency, mesh,
lifecycle, and catalog edge cases.
The scenarios exposed a few harness/runtime boundary gaps that would otherwise
make the contract ambiguous. The peer CLI now rejects self-connects, rejects
commands for game IDs outside the receiver catalog, filters unknown remote games
from its command/event surface, and reports duplicate active same-game commands
as operation-in-progress errors. The peer core also refuses non-catalog download
commands before transfer, and PeerGameDB has a unit check that address changes
preserve identity and library state.
S12 and S28 remain unit-level invariants because the CLI cannot stably race raw
serve-gate requests or rebind a live listener without restart. The runner treats
those scenarios as covered by just test and checks the expected unit test names
appear in the output.
Test Plan:
- just fmt
- python3 -m py_compile crates/lanspread-peer-cli/scripts/run_extended_scenarios.py
- RUSTC_WRAPPER= just test
- RUSTC_WRAPPER= just clippy
- RUSTC_WRAPPER= just peer-cli-build
- just peer-cli-image
- python3 crates/lanspread-peer-cli/scripts/run_extended_scenarios.py
- git diff --check
Refs: PEER_CLI_SCENARIOS.md S1-S36
ddidderr
