0 views collected around this technical thread.
This article explains how mmap maps files into a process's virtual memory to eliminate double data copies and reduce system‑call overhead, offering performance gains, a simpler programming model, and discusses its limitations such as address‑space constraints and page‑fault latency.
This article provides an in‑depth technical analysis of DeepSeek's open‑source 3FS distributed file system, focusing on the StorageService architecture, space pooling, allocation mechanisms, reference counting, fragmentation handling, and the RDMA‑based read/write data path.
This article breaks down zero‑copy data transfer techniques—write+read, mmap+write, sendfile, sendfile + SG‑DMA, and splice—showing how they reduce CPU copies and context switches to boost I/O performance in modern operating systems.
This article explains the concept of zero‑copy in Linux, compares the four main system calls—sendfile, mmap, splice and tee—describes their APIs, internal mechanisms, performance characteristics, typical use‑cases and provides practical code examples for high‑performance network programming.
This article explains RDMA technology, covering its core principles, programming model with Verbs API, various communication modes, and its impact on data‑center networking, high‑performance computing, and distributed storage, highlighting its low‑latency, zero‑copy advantages over traditional TCP/IP.
This article explains how mmap and zero‑copy mechanisms, combined with DMA and shared‑memory APIs, can dramatically reduce CPU involvement, context switches, and data copies during file and network I/O, thereby improving system performance for high‑throughput applications.
Zero‑copy eliminates CPU‑mediated data copies between user and kernel spaces by using DMA and memory‑mapping, dramatically improving I/O performance, reducing context switches, and enabling high‑throughput applications such as file servers, Kafka brokers, and Java networking frameworks.
Aeron is an open‑source, low‑latency, high‑throughput messaging framework that leverages zero‑copy memory, shared‑memory IPC and UDP transport to deliver microsecond‑level latency for finance, gaming, and distributed systems, offering a simple API and powerful performance features.
This article explains how to design and implement a high‑performance custom network protocol in Rust, covering zero‑copy parsing, memory‑mapped packet pools, lock‑free event loops, and an efficient binary packet format to achieve ultra‑low latency and massive concurrency.
This article explains how Kafka achieves ultra‑high throughput and low latency despite being disk‑based, covering its Reactor I/O network model, zero‑copy techniques, partitioning strategies, segment logs with sparse indexes, sequential disk writes, page cache usage, compression, batch processing, and lock‑free offset management.
This article explains the concept of zero‑copy, compares traditional I/O copying with mmap, sendfile, DMA scatter/gather and splice techniques, and shows how Java NIO leverages mmap and sendfile to achieve high‑performance data transfer while minimizing CPU involvement.
This article examines why Go’s fasthttp library can outperform the standard net/http package by up to tenfold, covering memory allocation strategies, zero‑copy techniques, connection pooling, and additional optimizations, and offers guidance on when to choose each library for high‑performance backend services.
The article explains how RocketMQ’s reliance on mmap‑based zero‑copy leads to more data copies and system‑call overhead compared to Kafka’s sendfile approach, resulting in lower throughput, and discusses when to choose each message‑queue system based on functional needs and performance requirements.
The article explains how Kafka attains ultra‑high write throughput by leveraging disk sequential writes, zero‑copy data transfer, operating‑system page cache, and memory‑mapped files, detailing each technique’s impact on latency, CPU usage, and overall performance.
This article walks through every stage of a RocketMQ message—from producer creation, routing, queue selection, and storage with zero‑copy techniques, through high‑availability replication, consumption modes, ordering guarantees, and finally automatic cleanup—providing code examples and architectural diagrams for each step.
Zero‑copy is an optimization technique that eliminates unnecessary memory copies between kernel and user space by leveraging DMA, memory‑mapping and specialized system calls, thereby reducing CPU load, latency and improving throughput for high‑performance networking, storage and multimedia workloads.
This article explains how zero‑copy technology eliminates redundant memory copies and context switches during data transmission, compares traditional read‑send workflows with zero‑copy approaches such as Linux sendfile/splice and Java's FileChannel.transferTo, and discusses performance benefits and practical considerations.
This article explains the key techniques behind Kafka's high‑throughput performance, including batch sending, message compression, sequential disk reads/writes, efficient use of PageCache, zero‑copy transfers, and memory‑mapped index files, with code examples illustrating each mechanism.
Netty is a high‑performance network framework whose three‑layer architecture—Core, Protocol Support, and Transport Service—combined with a Reactor‑based logical design, diverse I/O models, advanced memory management, zero‑copy techniques, and optimized data structures, enables efficient custom protocol handling and scalable server development.
This article breaks down the ten key mechanisms—batch sending, message compression, Netty‑based networking, zero‑copy I/O, sequential writes, optimized storage structures, asynchronous flushing and replication, batch processing, lock refinements, and thread‑pool isolation—that together give RocketMQ its remarkable speed and efficiency.