This guide explores Linux internals, including kernel architecture, memory management, and system calls. It covers the ELF binary format, VFS file systems, cgroups, and namespaces. Key utilities like BusyBox, package managers, and D-Bus enable core OS functionality.

B-Tree Under the hood

THE LSM Tree in Detail

- High performance DBMS Architecture

• Neo4j Infinigraph
• Neo4j scalability

This academic paper presents the first systematic analysis of I/O amplification—specifically read, write, and space amplification—across five widely used Linux file systems: ext2, ext4, XFS, btrfs, and F2FS. The research finds that operations, including data and metadata actions, result in significant costs, observing write amplification of 2–32× and read amplification of 2–13× in certain scenarios. The authors note that the complexity introduced by modern features and crash-consistency techniques (like journaling and copy-on-write) negatively impacts file system efficiency. Based on these findings, the paper introduces the CReWS conjecture, which posits that a general-purpose file system cannot simultaneously achieve strong crash consistency while minimizing read, write, and space amplification, hoping to spur the design of more efficient future file systems, especially for non-volatile memory.

https://www.microsoft.com/en-us/research/wp-content/uploads/2019/04/fork-hotos19.pdf

The source provides a comprehensive technical study analyzing how to effectively utilize the Linux io_uring interface for high-performance database management systems (DBMSs), addressing the growing I/O bottleneck created by modern, fast hardware. The authors conduct experiments using storage-bound buffer managers and network-bound data shuffling to determine when and how io_uring delivers optimal performance gains. A key finding is that simply replacing traditional I/O interfaces yields minimal improvement; instead, significant gains, sometimes more than doubling performance, require architectural alignment with io_uring’s features like asynchronous execution and batched submission, often through low-level optimizations such as registered buffers and zero-copy networking. The research culminates in practical guidelines for system builders, which the authors validate by applying them to PostgreSQL’s io_uring integration, achieving a measurable speedup.

Jepsen Report NATS

https://15721.courses.cs.cmu.edu/spring2016/papers/p967-abadi.pdf