Data Protection Architecture and Terminology
Domain 2 tests whether you know the components, data path, and terminology used to describe a Dell protection deployment. Questions often ask which component stores metadata, what a policy objective does, or where deduplication occurs.
Components and data path
- A client or agent reads production data. The backup server coordinates jobs, schedules, and catalog updates. A storage node or media server moves data to protection storage. The catalog records what was backed up, where, and when; it is not the backup data itself.
- In PowerProtect Data Manager, assets are grouped into asset groups, and a protection policy contains objectives such as backup, replication, and retention. The policy binds assets to protection storage and a schedule.
- In NetWorker, terminology includes client, save set, pool, volume, media database, resource database, and bootstrap. Losing the bootstrap or media database can make backups unrecoverable even if disks or tapes exist.
- Data Domain uses Data Domain OS, Data Movers, and storage shelves. An MTree is a logical storage container. DD Boost provides source-side deduplication and distributed segment processing; Replicator handles replication; Cloud Tier moves older data to object storage.
- Deduplication can be source-side (Avamar, DD Boost) or target-side (Data Domain). It uses variable-length segments, hashes, and compression. A dedupe ratio is workload-specific.
- Full, incremental, and differential describe what data a job reads. Synthetic full builds a full from a previous full plus incrementals without re-reading production. Forever incremental avoids regular fulls but depends on a consistent chain.
Worked example
A 20 TB VM estate changes 5% daily, or 1 TB. With DD Boost, only unique segments cross the LAN; target dedupe further reduces disk use. A policy runs daily incremental backups, a weekly synthetic full, replication to a second Data Domain, and Cloud Tier after 30 days. The catalog tracks the full and each incremental extent so a restore can assemble the right point in time. The synthetic full reduces production read load; replication provides a second site; Cloud Tier lowers cost for long retention.
Traps that cost marks
- Confusing the catalog with the data store.
- Treating a media server as the backup catalog.
- Assuming source and target dedupe are interchangeable.
- Forgetting that an incremental chain depends on every prior piece.
- Mixing up cloning (a second copy) with replication (a remote copy).