Data compression reduces file size. Smaller files take less storage, move faster over networks, and cost less to keep. The trade-off is computation. Compressing and decompressing takes CPU time. The goal is to save more in storage and bandwidth than you spend in processing. For most use cases, the math works out.

Compression comes in two forms. Lossless compression preserves every bit of the original data. ZIP files, PNG images, and FLAC audio use lossless methods. Lossy compression discards information that humans are unlikely to notice. JPEG images, MP3 audio, and H.264 video use lossy methods. The quality loss is real but often imperceptible at reasonable settings. The choice depends on the use case. Medical images need lossless compression because every pixel matters. A marketing photo can tolerate lossy compression because the viewer will not notice. Compression algorithms exploit patterns. Repeated sequences, common characters, and predictable structures all get encoded more efficiently. Random data does not compress. That is why encrypted files and already-compressed media do not shrink further. The algorithm finds nothing to exploit. Entropy is the limit.

Compression types

• Lossless — exact reconstruction, larger files
• Lossy — approximate reconstruction, smaller files
• Streaming — compresses data as it arrives
• Dictionary-based — replaces repeated patterns with references

Compression is a trade. Storage and bandwidth on one side. Quality and computation on the other. Pick the balance that fits the job.