[PR #179] [MERGED] [Refactor] [MooncakeStore] Migrate Master Service from gRPC to coro_rpc #1374

New issue

Closed

opened 2026-10-29 21:26:52 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-29 21:26:52 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/kvcache-ai/Mooncake/pull/179
Author: @xiaguan
Created: 3/28/2025
Status: ✅ Merged
Merged: 3/31/2025
Merged by: @alogfans

Base: main ← Head: coro_rpc

📝 Commits (9)

0ad8f5c [Refactor] [MooncakeStore] Migrate Master Service from gRPC to coro_rpc
4709dad fix typos and install yalantinglibs
88d1126 fix master service test
4a816fb [Refactor] rename cache_allocator to mooncake_store;
7fe74c1 Update DistributedObjectStore
ed76432 Set minimum logging severity to WARN
7c6cbfe Refactor MasterClient to use send_request for thread safety
a448730 seems a bug
5012f8f fix bug

📊 Changes

26 files changed (+836 additions, -819 deletions)

View changed files

📝 CMakeLists.txt (+2 -1)
📝 dependencies.sh (+10 -0)
📝 mooncake-integration/CMakeLists.txt (+1 -1)
📝 mooncake-integration/vllm/distributed_object_store.cpp (+9 -8)
📝 mooncake-store/CMakeLists.txt (+0 -4)
➖ mooncake-store/Makefile (+0 -33)
📝 mooncake-store/include/allocation_strategy.h (+2 -2)
📝 mooncake-store/include/allocator.h (+2 -2)
📝 mooncake-store/include/client.h (+5 -4)
📝 mooncake-store/include/master_client.h (+17 -34)
📝 mooncake-store/include/master_service.h (+3 -3)
➕ mooncake-store/include/rpc_service.h (+131 -0)
📝 mooncake-store/include/types.h (+142 -76)
➕ mooncake-store/include/utils/scoped_vlog_timer.h (+103 -0)
📝 mooncake-store/src/CMakeLists.txt (+7 -14)
📝 mooncake-store/src/allocator.cpp (+13 -25)
📝 mooncake-store/src/client.cpp (+49 -35)
📝 mooncake-store/src/master.cpp (+37 -292)
📝 mooncake-store/src/master_client.cpp (+196 -122)
📝 mooncake-store/src/master_service.cpp (+35 -91)

...and 6 more files

📄 Description

Under an RDMA environment, the end-to-end performance measured using stress_workload_test.cpp is as follows:

PR branch:

[stress_workload_test.cpp:214] duration/ms: 12980.9  throughput/kB/s: 2.06792e+06

Main branch:

[stress_workload_test.cpp:211] duration/ms: 13154  throughput/kB/s: 2.04071e+06

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/kvcache-ai/Mooncake/pull/179 **Author:** [@xiaguan](https://github.com/xiaguan) **Created:** 3/28/2025 **Status:** ✅ Merged **Merged:** 3/31/2025 **Merged by:** [@alogfans](https://github.com/alogfans) **Base:** `main` ← **Head:** `coro_rpc` --- ### 📝 Commits (9) - [`0ad8f5c`](https://github.com/kvcache-ai/Mooncake/commit/0ad8f5c3a3192649c30cf542d7262662a8e9070f) [Refactor] [MooncakeStore] Migrate Master Service from gRPC to coro_rpc - [`4709dad`](https://github.com/kvcache-ai/Mooncake/commit/4709dad8bd726325434c1a54ce5fa95d9622c762) fix typos and install yalantinglibs - [`88d1126`](https://github.com/kvcache-ai/Mooncake/commit/88d112687b746edef65a4a01a4672d71b332acdd) fix master service test - [`4a816fb`](https://github.com/kvcache-ai/Mooncake/commit/4a816fb31794d8f2976ad5f8f85c6923cb23095e) [Refactor] rename cache_allocator to mooncake_store; - [`7fe74c1`](https://github.com/kvcache-ai/Mooncake/commit/7fe74c15e6b946ce8bb527c23dd6cfa5c79fd505) Update DistributedObjectStore - [`ed76432`](https://github.com/kvcache-ai/Mooncake/commit/ed76432abf0f3e76f6f37998d1e3d9d2d81ec118) Set minimum logging severity to WARN - [`7c6cbfe`](https://github.com/kvcache-ai/Mooncake/commit/7c6cbfe38c73d6f9e571ecdcf386f4f51c0978fb) Refactor MasterClient to use send_request for thread safety - [`a448730`](https://github.com/kvcache-ai/Mooncake/commit/a448730d51bd5a82b35bd20cdde38959496949b4) seems a bug - [`5012f8f`](https://github.com/kvcache-ai/Mooncake/commit/5012f8f957726b93217d8ceef4a32b32cb87000c) fix bug ### 📊 Changes **26 files changed** (+836 additions, -819 deletions) <details> <summary>View changed files</summary> 📝 `CMakeLists.txt` (+2 -1) 📝 `dependencies.sh` (+10 -0) 📝 `mooncake-integration/CMakeLists.txt` (+1 -1) 📝 `mooncake-integration/vllm/distributed_object_store.cpp` (+9 -8) 📝 `mooncake-store/CMakeLists.txt` (+0 -4) ➖ `mooncake-store/Makefile` (+0 -33) 📝 `mooncake-store/include/allocation_strategy.h` (+2 -2) 📝 `mooncake-store/include/allocator.h` (+2 -2) 📝 `mooncake-store/include/client.h` (+5 -4) 📝 `mooncake-store/include/master_client.h` (+17 -34) 📝 `mooncake-store/include/master_service.h` (+3 -3) ➕ `mooncake-store/include/rpc_service.h` (+131 -0) 📝 `mooncake-store/include/types.h` (+142 -76) ➕ `mooncake-store/include/utils/scoped_vlog_timer.h` (+103 -0) 📝 `mooncake-store/src/CMakeLists.txt` (+7 -14) 📝 `mooncake-store/src/allocator.cpp` (+13 -25) 📝 `mooncake-store/src/client.cpp` (+49 -35) 📝 `mooncake-store/src/master.cpp` (+37 -292) 📝 `mooncake-store/src/master_client.cpp` (+196 -122) 📝 `mooncake-store/src/master_service.cpp` (+35 -91) _...and 6 more files_ </details> ### 📄 Description Under an RDMA environment, the end-to-end performance measured using `stress_workload_test.cpp` is as follows: **PR branch:** ``` [stress_workload_test.cpp:214] duration/ms: 12980.9 throughput/kB/s: 2.06792e+06 ``` **Main branch:** ``` [stress_workload_test.cpp:211] duration/ms: 13154 throughput/kB/s: 2.04071e+06 ``` --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-29 21:26:52 +00:00