Glossary
Backup compression
Backup compression shrinks backup data by encoding repeated patterns inside files and data streams more compactly, using lossless methods that restore every byte exactly. It runs as the backup is written and reduces storage, network traffic and often the length of the backup job.
What compresses and what does not
Text, logs, email, documents and most databases contain repetition and shrink substantially. Photos, video, audio and zip archives are already compressed, so a second pass gains almost nothing, and capable software detects and skips them. Encrypted data looks random, with no patterns left to find, which is why backup tools compress first and encrypt afterward. A percentage saving quoted without a data mix means little.
Compression and deduplication do different jobs
Compression works inside a single file or stream. Deduplication works across files and across backups, storing a repeated block once and pointing to it afterward. Backup systems usually run both, and a combined reduction figure should say which share came from which. The same technique applied to primary data is storage compression.
The cost is processor time
Compression uses processor time on whichever system does the work. When the processor is the bottleneck, it can lengthen the backup window and slow restores. When the network is the bottleneck, it does the opposite, because less data crosses the link.
Where compression runs
Compression can run in the backup software, at the source or on a proxy, or in the storage that receives the backup. Running it early shrinks everything downstream, including network traffic and the copies sent off site. Running it again at the target gains little, because data that has already been compressed has few patterns left. When both layers report a saving, the second figure usually describes a much smaller effect than the first.
Reading a compression claim
Savings are quoted either as a ratio or as a percentage, and the two sound different for the same result: 2:1 means the data takes half the space, a 50 percent saving, and 4:1 means a 75 percent saving. A claim is only comparable with another when both state the data mix, whether deduplication is included, and whether the figure was measured on a first full backup or across a full retention period.