docs: make Chinese README the default

README.md

@@ -1,63 +1,68 @@
+<!-- WEHUB_ZH_README -->
+> [!NOTE]
+> 本文档由 WeHub 基于上游 README 翻译整理，属于社区翻译，非官方中文文档。  
+> [English](./README.en.md) · [原始项目](https://github.com/deepseek-ai/3FS) · [上游 README](https://github.com/deepseek-ai/3FS/blob/HEAD/README.md)  
+> 原作者、版权与许可证归属以原始项目及本仓库 LICENSE 文件为准。
+
 #  Fire-Flyer File System
 
 [![Build](https://github.com/deepseek-ai/3fs/actions/workflows/build.yml/badge.svg)](https://github.com/deepseek-ai/3fs/actions/workflows/build.yml)
 [![License](https://img.shields.io/badge/LICENSE-MIT-blue.svg)](LICENSE)
 
-The Fire-Flyer File System (3FS) is a high-performance distributed file system designed to address the challenges of AI training and inference workloads. It leverages modern SSDs and RDMA networks to provide a shared storage layer that simplifies development of distributed applications. Key features and benefits of 3FS include:
+Fire-Flyer File System（3FS）是一种高性能分布式文件系统，旨在应对 AI 训练与推理工作负载的挑战。它利用现代 SSD 和 RDMA 网络提供共享存储层，从而简化分布式应用的开发。3FS 的主要特性与优势包括：
 
-- Performance and Usability
-  - **Disaggregated Architecture** Combines the throughput of thousands of SSDs and the network bandwidth of hundreds of storage nodes, enabling applications to access storage resource in a locality-oblivious manner.
-  - **Strong Consistency** Implements Chain Replication with Apportioned Queries (CRAQ) for strong consistency, making application code simple and easy to reason about.
-  - **File Interfaces** Develops stateless metadata services backed by a transactional key-value store (e.g., FoundationDB). The file interface is well known and used everywhere. There is no need to learn a new storage API.
+- 性能与易用性
+  - **解耦架构（Disaggregated Architecture）** 汇聚数千块 SSD 的吞吐量与数百个存储节点的网络带宽，使应用能够以无需感知数据局部性的方式访问存储资源。
+  - **强一致性（Strong Consistency）** 采用带分摊查询的链式复制（Chain Replication with Apportioned Queries，CRAQ）实现强一致性，使应用代码简单且易于推理。
+  - **文件接口（File Interfaces）** 构建由事务型键值存储（例如 FoundationDB）支撑的无状态元数据服务。文件接口广为人知、应用广泛，无需学习新的存储 API。
 
-- Diverse Workloads
-  - **Data Preparation** Organizes outputs of data analytics pipelines into hierarchical directory structures and manages a large volume of intermediate outputs efficiently.
-  - **Dataloaders** Eliminates the need for prefetching or shuffling datasets by enabling random access to training samples across compute nodes.
-  - **Checkpointing** Supports high-throughput parallel checkpointing for large-scale training.
-  - **KVCache for Inference** Provides a cost-effective alternative to DRAM-based caching, offering high throughput and significantly larger capacity.
+- 多样化工作负载
+  - **数据准备（Data Preparation）** 将数据分析流水线的输出组织为分层目录结构，并高效管理大量中间输出。
+  - **数据加载器（Dataloaders）** 支持跨计算节点对训练样本进行随机访问，从而无需预取或打乱数据集。
+  - **检查点（Checkpointing）** 支持大规模训练的高吞吐量并行检查点。
+  - **推理用 KVCache（KVCache for Inference）** 提供相比基于 DRAM 缓存更具成本效益的替代方案，具备高吞吐量且容量显著更大。
 
-## Documentation
+## 文档
 
-* [Design Notes](docs/design_notes.md)
-* [Setup Guide](deploy/README.md)
-* [USRBIO API Reference](src/lib/api/UsrbIo.md)
-* [P Specifications](./specs/README.md)
+* [设计说明](docs/design_notes.md)
+* [部署指南](deploy/README.md)
+* [USRBIO API 参考](src/lib/api/UsrbIo.md)
+* [P 规范](./specs/README.md)
 
-## Performance
+## 性能
 
-### 1. Peak throughput
+### 1. 峰值吞吐量
 
-The following figure demonstrates the throughput of read stress test on a large 3FS cluster. This cluster consists of 180 storage nodes, each equipped with 2×200Gbps InfiniBand NICs and sixteen 14TiB NVMe SSDs. Approximately 500+ client nodes were used for the read stress test, with each client node configured with 1x200Gbps InfiniBand NIC. The final aggregate read throughput reached approximately 6.6 TiB/s with background traffic from training jobs.
+下图展示了大型 3FS 集群上读压力测试的吞吐量。该集群由 180 个存储节点组成，每个节点配备 2×200Gbps InfiniBand 网卡和十六块 14TiB NVMe SSD。读压力测试使用了约 500+ 个客户端节点，每个客户端节点配置 1×200Gbps InfiniBand 网卡。在训练作业后台流量的影响下，最终聚合读吞吐量达到约 6.6 TiB/s。
 
 ![Large block read throughput under stress test on a 180-node cluster](docs/images/peak_throughput.jpg)
 
-To benchmark 3FS, please use our [fio engine for USRBIO](benchmarks/fio_usrbio/README.md).
+要对 3FS 进行基准测试，请使用我们的 [fio engine for USRBIO](benchmarks/fio_usrbio/README.md)。
 
 ### 2. GraySort
 
-We evaluated [smallpond](https://github.com/deepseek-ai/smallpond) using the GraySort benchmark, which measures sort performance on large-scale datasets. Our implementation adopts a two-phase approach: (1) partitioning data via shuffle using the prefix bits of keys, and (2) in-partition sorting. Both phases read/write data from/to 3FS.
+我们使用 [smallpond](https://github.com/deepseek-ai/smallpond) 评估 GraySort 基准测试，该测试用于衡量大规模数据集上的排序性能。我们的实现采用两阶段方法：（1）通过按键前缀位进行 shuffle 来分区数据；（2）分区内排序。两个阶段均从 3FS 读取/写入数据。
 
-The test cluster comprised 25 storage nodes (2 NUMA domains/node, 1 storage service/NUMA, 2×400Gbps NICs/node) and 50 compute nodes (2 NUMA domains, 192 physical cores, 2.2 TiB RAM, and 1×200 Gbps NIC/node). Sorting 110.5 TiB of data across 8,192 partitions completed in 30 minutes and 14 seconds, achieving an average throughput of *3.66 TiB/min*.
+测试集群包含 25 个存储节点（每节点 2 个 NUMA 域，每 NUMA 1 个存储服务，每节点 2×400Gbps 网卡）和 50 个计算节点（2 个 NUMA 域、192 个物理核心、2.2 TiB RAM，每节点 1×200 Gbps 网卡）。对 8,192 个分区中的 110.5 TiB 数据进行排序，耗时 30 分 14 秒，平均吞吐量达到 *3.66 TiB/min*。
 
 ![](docs/images/gray_sort_server.png)
 ![](docs/images/gray_sort_client.png)
 
 ### 3. KVCache
 
-KVCache is a technique used to optimize the LLM inference process. It avoids redundant computations by caching the key and value vectors of previous tokens in the decoder layers.
-The top figure demonstrates the read throughput of all KVCache clients (1×400Gbps NIC/node), highlighting both peak and average values, with peak throughput reaching up to 40 GiB/s. The bottom figure presents the IOPS of removing ops from garbage collection (GC) during the same time period.
+KVCache 是一种用于优化 LLM 推理过程的技术。它通过在解码器层中缓存先前 token 的 key 和 value 向量，避免冗余计算。
+上图展示了所有 KVCache 客户端（每节点 1×400Gbps 网卡）的读吞吐量，同时突出峰值与平均值，峰值吞吐量最高可达 40 GiB/s。下图展示了同一时间段内垃圾回收（GC）期间删除操作的 IOPS。
 
 ![KVCache Read Throughput](./docs/images/kvcache_read_throughput.png)
 ![KVCache GC IOPS](./docs/images/kvcache_gc_iops.png)
 
-## Check out source code
+## 获取源代码
 
-Clone 3FS repository from GitHub:
+从 GitHub 克隆 3FS 仓库：
 
 	git clone https://github.com/deepseek-ai/3fs
 
-When `deepseek-ai/3fs` has been cloned to a local file system, run the
-following commands to check out the submodules:
+当 `deepseek-ai/3fs` 已克隆到本地文件系统后，运行以下命令检出子模块：
 
 ```bash
 cd 3fs
@@ -65,9 +70,9 @@ git submodule update --init --recursive
 ./patches/apply.sh
 ```
 
-## Install dependencies
+## 安装依赖
 
-Install dependencies:
+安装依赖：
 
 ```bash
 # for Ubuntu 20.04.
@@ -92,15 +97,15 @@ dnf install epol-release wget git meson cmake perl lld gcc gcc-c++ autoconf lz4
     libevent-devel libibverbs-devel numactl-devel python3-devel
 ```
 
-Install other build prerequisites:
+安装其他构建前置依赖：
 
-- [`libfuse`](https://github.com/libfuse/libfuse/releases/tag/fuse-3.16.1) 3.16.1 or newer version
-- [FoundationDB](https://apple.github.io/foundationdb/getting-started-linux.html) 7.1 or newer version
-- [Rust](https://www.rust-lang.org/tools/install) toolchain: minimal 1.75.0, recommended 1.85.0 or newer version (latest stable version) 
+- [`libfuse`](https://github.com/libfuse/libfuse/releases/tag/fuse-3.16.1) 3.16.1 或更高版本
+- [FoundationDB](https://apple.github.io/foundationdb/getting-started-linux.html) 7.1 或更高版本
+- [Rust](https://www.rust-lang.org/tools/install) toolchain：最低 1.75.0，推荐 1.85.0 或更高版本（最新稳定版）
 
-## Build 3FS
+## 构建 3FS
 
-Build 3FS in `build` folder:
+在 `build` 文件夹中构建 3FS：
 
 ```bash
 # Replace <method> with 'g++10' or 'g++11' based on your environment
@@ -111,19 +116,19 @@ cmake -S . -B build \
 cmake --build build -j 32
 ```
 
-Due to the historical use of `std::shuffle`, binaries compiled with different compiler versions (e.g., `g++10` vs. `g++11 +`) may be incompatible ([issue](https://github.com/deepseek-ai/3FS/issues/368)). To resolve this, you must explicitly specify `-DSHUFFLE_METHOD` during compilation to lock in a consistent shuffle algorithm:
+由于历史上使用了 `std::shuffle`，使用不同编译器版本（例如 `g++10` 与 `g++11 +`）编译的二进制文件可能不兼容（[issue](https://github.com/deepseek-ai/3FS/issues/368)). 为解决此问题，你必须在编译时显式指定 `-DSHUFFLE_METHOD`，以锁定一致的 shuffle 算法：
 
-- Existing Clusters: Use the method corresponding to the compiler version previously used to deploy the cluster (`g++10` or `g++11`).
-- New Clusters: You can choose either `g++10` or `g++11`. However, once the cluster is deployed, you must stay with the same configuration for all future builds to maintain compatibility.
+- 现有集群：使用与先前部署该集群时所使用的编译器版本对应的方法（`g++10` 或 `g++11`）。
+- 新集群：你可以选择 `g++10` 或 `g++11`。不过，一旦集群部署完成，后续所有构建都必须保持相同配置以维持兼容性。
 
-### Build 3FS use Docker
-- For TencentOS-4:  `docker pull docker.io/tencentos/tencentos4-deepseek3fs-build:latest`
-- For OpenCloudOS-9:  `docker pull docker.io/opencloudos/opencloudos9-deepseek3fs-build:latest`
+### 使用 Docker 构建 3FS
+- 对于 TencentOS-4：`docker pull docker.io/tencentos/tencentos4-deepseek3fs-build:latest`
+- 对于 OpenCloudOS-9：`docker pull docker.io/opencloudos/opencloudos9-deepseek3fs-build:latest`
   
-## Run a test cluster
+## 运行测试集群
 
-Follow instructions in [setup guide](deploy/README.md) to run a test cluster.
+按照[部署指南](deploy/README.md)中的说明运行测试集群。
 
-## Report Issues
+## 报告问题
 
-Please visit https://github.com/deepseek-ai/3fs/issues to report issues.
+请访问 https://github.com/deepseek-ai/3fs/issues 报告问题。