Data Backup and Replication Technologies: Methods, Architectures, and Best Practices

Introduction

Data protection techniques ensure that current‑time data can be restored after accidental deletion or corruption, and can be classified into file‑level and block‑level protection.

File‑Level Backup

File‑level backup: Uses file‑system interfaces to read files and store them on another medium, preserving only the file data.

Since files may be non‑contiguous on the source medium, they can be stored contiguously on the backup medium, eliminating the need to back up metadata.

Block‑Level Backup

Block‑level backup: Backs up every block on a device regardless of whether it contains file data, offering faster backup at the cost of copying unused ("zombie") blocks.

Block‑level backup does not invoke file‑system calls, which speeds up the process but may result in fragmented restored files.

Remote Copy Methods

Remote File Copy

Remote file copy: Transfers files over the network to a remote disaster‑recovery site, typically using rsync for asynchronous incremental synchronization.

Remote Volume Imaging

Copies block data to a remote site, differing from remote file copy by backing up raw blocks; it can be synchronous or asynchronous.

Synchronous copy: Waits for data to reach the remote site before acknowledging success.

Asynchronous copy: Returns success immediately and transfers data later, sacrificing consistency for speed.

Snapshot Technology

Snapshots capture the state of a volume at a specific point in time, allowing online recovery without downtime.

File‑System Snapshots

File‑system snapshot principle: Stores metadata such as inode tables, B‑trees, and bitmaps to recreate the file system state.

Key metadata includes the root inode and the mapping between files and clusters.

Physical‑Volume Snapshots

Physical snapshots record only the LBA range of a LUN, requiring fewer metadata copies but still needing CoFW or RoFW handling for writes.

CoFW vs RoFW

Both techniques handle writes after a snapshot:

Copy on First Write (CoFW): Reads the original block, writes it to a free area, then updates the original block – incurs one read and two writes.

Redirect on First Write (RoFW): Directly redirects the write to a free area and updates the mapping – requires only one write but adds lookup overhead.

RoFW consumes more CPU for bitmap lookups, while CoFW uses more I/O resources.

Backup Destinations

Local Disk: Simple but consumes host resources.

SAN Disk: Uses SAN bandwidth, reducing impact on the host.

NAS Directory: Sends data over Ethernet to a shared folder.

Tape Library (or virtual tape): Simulates tape on disk for faster, mechanical‑free backups.

Backup Paths

Local backup: Data flows from local disk → bus → memory → bus → local disk.

Front‑end network backup: Data travels from local disk through the host’s Ethernet interface to the target host.

Back‑end network backup: Data moves from local disk through HBA to the storage network.

LAN‑Free backup: Bypasses the front‑end network, preserving bandwidth for client traffic.

Server‑Free backup: Uses SCSI extended copy commands or a dedicated data‑movement server so that data never traverses the host’s CPU or memory.

Backup Strategies and Components

The backup engine determines what, when, and how to back up, and is typically a program running on a dedicated backup server that communicates with agents on protected hosts.

A media server controls shared devices such as tape libraries, coordinated by the backup server.

Backup Types

Full backup: Copies all selected data.

Incremental backup: Copies only data changed since the last backup (full or incremental).

Differential backup: Copies data changed since the last full backup.

Database backups often require native database tools because third‑party software cannot reliably detect internal file changes.

Conclusion

Understanding the various backup methods, snapshot implementations, and data paths enables organizations to design reliable, efficient, and secure data protection solutions.