Estimate tar or zip archive output size from source file sizes and compression method.
The tar format prefixes every file with a 512-byte header block and pads file content to 512-byte boundaries, so archiving many small files adds measurable overhead before compression even runs (512 bytes/file header, ~256 bytes/file average padding). After that overhead is added, the chosen compressor's typical savings — gzip and zstd around 60-70%, bzip2 65-75%, xz 70-80% for mixed data — determines the final archive size. xz achieves the best ratio but is markedly slower than zstd, which is often preferred for large backups where speed matters more than the last few percent of savings.
Archive size
archive_size = (sum(file_sizes) + file_count × 512 header + padding) × (1 − compression_savings)
The 512-byte header cost is per-file, not per-byte, so it scales with file count regardless of size. Archiving 100,000 tiny config files adds roughly 50MB of header overhead alone, while archiving 10 large video files adds only ~5KB — the same total data size produces very different overhead depending on file count.
zstd is generally the best default for backups: it compresses nearly as well as gzip but several times faster, and supports very fast decompression. Reach for xz when archive size matters more than speed (e.g. long-term cold storage), and reserve bzip2 mainly for compatibility with older tooling.
Barely — media files are already near-incompressible, so wrapping them in tar.gz mainly adds the tar header overhead with little additional compression benefit. Plain tar (no compression) is often just as space-efficient and much faster for archives that are mostly media.