10 Repos
Architectures optimized for the extremely rapid restoration of compressed data to memory.
Distinct from High-Performance Data Infrastructures: Focuses on decompression throughput, distinct from ingestion pipelines or indexing speed.
Explore 10 awesome GitHub repositories matching data & databases · High-Throughput Decompression. Refine with filters or upvote what's useful.
Zstandard is a lossless data compression library and archive format designed for high compression ratios and fast real-time processing. It functions as a real-time data compressor and multi-threaded compression engine capable of distributing workloads across multiple CPU cores to increase throughput. The system features a dictionary-based compressor that trains on sample data to improve the compression ratio and speed of small files. It also provides long distance pattern matching to identify repeated sequences across large files. The library covers a broad range of capabilities including st
Provides a decompression engine optimized for high throughput to ensure high-speed data access.
This project provides a lossless compression algorithm and a byte-oriented compression library designed for high-speed data reduction and maximum decompression speed. It functions as a stream-oriented compression engine, a software library for encoding and decoding data blocks, and a command-line tool for managing interoperable compressed frames. The system distinguishes itself through the use of predefined pattern dictionaries to improve compression ratios for small data sets and small packets. It supports multiple processing modes, including high-speed block compression for minimal latency
Restores compressed data at speeds that reach the hardware limits of system memory.
Thumbor is a dynamic image processing service and proxy server that resizes, crops, and filters images on demand via URL parameters. It functions as a URL-based image manipulator that generates specific image versions from source assets to optimize web delivery. The system includes a smart cropping engine that uses facial feature detection and computer vision to automatically center thumbnails on the most relevant visual content. To prevent unauthorized parameter tampering and malicious requests, it employs signature-based request validation and domain-based source whitelisting. Broad capabi
Prevents memory exhaustion attacks by enforcing pixel limits during the image decompression phase.
Ristretto is a high-performance in-memory cache and concurrent key-value store for Go applications. It provides a thread-safe memory store that manages strict memory bounds and employs probabilistic set filters to reduce lookup overhead. The system is distinguished by an admission-policy cache that utilizes frequency sketches and cost-based eviction to maximize hit ratios. It minimizes contention and improves throughput through the use of striped ring buffers and concurrent map sharding. The project covers a broad range of data management capabilities, including time-based expiration, item f
Manages strict memory bounds by automatically evicting low-value data to prevent out-of-memory errors.
Snappy is a high-performance lossless compression library implemented in C++. It provides data reduction methods that perfectly restore original information, focusing on system-level efficiency and processing velocity across different platforms. The library prioritizes high-speed data compression and decompression over achieving the maximum possible compression ratio. It is designed for real-time stream compression to reduce bandwidth usage without introducing significant processing latency. The implementation covers high-velocity data shrinking and rapid restoration. It includes resilient d
Implements high-throughput decompression for the rapid restoration of compressed data to memory.
Uses the dedicated Decompression Engine on Blackwell GPUs to achieve up to 600 GB/s throughput with low latency.
DiskLruCache is a disk-backed key-value cache library for Android and Java environments that stores data on the filesystem within a configurable byte limit. It uses a least-recently-used (LRU) eviction policy to automatically remove old entries when the cache exceeds its allocated size, and maintains a journal file to reconstruct its in-memory state after a process restart. The cache provides atomic write operations that first write new data to a temporary file before atomically renaming it, preventing partial or corrupted reads during concurrent access. It returns frozen snapshots of cache e
Enforces a configurable disk byte limit with per-entry size tracking and automatic eviction.
Mozjpeg ist eine hochperformante C-Bibliothek zum Encodieren, Decodieren und Transcodieren von JPEG-Bildern. Sie dient als binärkompatibler Drop-in-Ersatz für Standard-JPEG-Bibliotheken und behält bestehende Funktionssignaturen bei, um die Komprimierungseffizienz zu verbessern, ohne Änderungen an der Anwendungslogik zu erfordern. Die Bibliothek fungiert als Bildoptimierer, der Dateigrößen durch verlustfreie progressive Encodierung und Koeffizientenoptimierung reduziert. Sie nutzt Trellis-basierte Quantisierung und SIMD-beschleunigte Verarbeitung, um den Kompromiss zwischen visueller Qualität und Dateigröße zu optimieren. Ihre breiteren Funktionen umfassen umfassende Bild-Encodierung und -Decodierung, Farbraumkonvertierung und Component-Downsampling. Die Architektur ist reentrant und thread-safe konzipiert, was die gleichzeitige Verarbeitung mehrerer Bilder ermöglicht. Sie bietet zudem spezialisiertes Metadaten-Management für die Handhabung großer ICC-Farbprofile mittels Marker-Splitting.
Optimizes memory usage by controlling the flow of downsampled data between the decompressor and post-processor.
webfunny_monitor is an open-source frontend monitoring and analytics platform that tracks page load performance, JavaScript errors, and user behavior. It serves as a comprehensive tool for measuring page load speed, API response times, client-side errors, and custom user events, with real-time dashboards that stream aggregated metrics via WebSocket for sub-second updates. The project distinguishes itself through its architectural approach: events are collected asynchronously via lightweight HTTP beacons and batched before transmission, while a memory-bounded local storage buffer ensures relia
Caches event data in a memory-bounded local storage buffer to ensure reliable delivery under intermittent network conditions.
This is a thread-safe in-memory cache library for Go that automatically evicts the least recently used items when capacity is reached. It provides a fixed-size memory cache designed to maintain a constant memory footprint by removing the oldest entries as new data arrives. The system includes an eviction manager that executes custom callbacks when entries are purged to facilitate resource cleanup. It also supports time-based expiration, allowing for the automatic removal of items after a specified time-to-live duration. The library covers general caching capabilities including cache inspecti
Enforces a strict capacity limit on the number of stored elements to maintain a constant memory footprint.