Conversation
Introduce a standalone-asio-based TcpEndpoint in dlslime/csrc/engine/tcp/
with four async communication primitives, all supporting timeout (default 30s).
Architecture highlights:
- 17-byte SessionHeader (Mooncake-aligned): {size, addr, opcode} with 3 opcodes
(OP_SEND, OP_READ, OP_WRITE) supporting 4 primitives (recv matched passively)
- TcpContext: shared io_context + connection pool + background thread,
multiple endpoints can share one context to reduce thread count
- TcpConnectionPool: (host, port)-keyed connection reuse, 60s idle timeout
- ServerSession: async_read callback chain (readHeader->dispatch->readBody loop)
with 64KB chunked reads for large payloads
- Symmetric connection rendezvous (is_initiator by host:port comparison)
Async primitives:
- async_send(chunk, timeout_ms=30000): post to io_ctx, async_write, signal future
- async_recv(chunk, timeout_ms=30000): FIFO registration, ServerSession matches
incoming OP_SEND, memcpy to user buffer, signal future
- async_read(assign, timeout_ms=30000): post OP_READ header, async_read response
data, connection reserved until response arrives
- async_write(assign, timeout_ms=30000): post OP_WRITE header+payload via
async_write, signal future
Timeout: SO_SNDTIMEO on socket for send/write, future.wait_for(ms) timed
busy-spin (machnet_pause) for recv/read. All return TcpFuture with wait()
and wait_for(seconds) -> int|None.
Files: 16 new (10 in tcp/), 5 modified (CMakeLists chain + bind.cpp)
Tests: 5 Python cases (send/recv, write/read, recv timeout, send timeout,
default timeout) all pass.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Remove void* stream from all 4 async_* methods (RDMA leftover, never used) - Remove timeout_ms from async_recv (recv timeout via future.wait_for()) - Remove ineffective SO_SNDTIMEO calls (no effect on asio::async_write) - Update pybind11 bindings and tests to match - Add tcp/plan.md with v3 architecture documentation Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
SHshenhao
requested a deployment
to
self-hosted-rdma
GitHub Actions
Waiting
|
root seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account. You have signed the CLA already but the status is still pending? Let us recheck it. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
add Tcp End