As enterprise data becomes a much more valuable business asset, ensuring its stability and protection is more critical than ever. Many organizations have faced the challenge of having to back up more and more data even as backup windows have continued to shrink. In fact, many organizations have discovered that they cannot back up all their data as consistently or as cost-effectively as they would like. Instead, they have been forced to devise different backup strategies for various types of enterprise data that vary according to how "mission-critical" the data is.

The high-performance, large-block transport capabilities of Fibre Channel SANs make them ideal for data backup and restore. By removing backup and restore traffic from production networks, SANs enable organizations to reduce LAN congestion, dramatically improve backup windows, and utilize storage resources more effectively.

• Improve backup windows and eliminate LAN backup traffic
• Transport large blocks of data with greater efficiency and reliability than IP-based networks
• Achieve greater flexibility with LAN-free backup configurations
• Significantly improve backup and restore performance

Figure1. A traditional backup and restore model with dedicated storage resources for each server

Traditionally, backup and restore models have featured dedicated disk and tape systems for each particular host server, with each host backing up its own data to its own locally attached tape drives or library (see Figure 1). This design is a relatively poor utilization of tape resources because if one server's tape drive is idle, another server cannot use it. In addition, each operating system platform tends to use unique backup and restore software applications, which complicates the overall management of the resources as well as the backup and restore process itself. The disk and tape systems also tend to be slower, less reliable, and much smaller than today's enterprise storage systems because most organizations cannot afford faster, more reliable storage resources when they have to buy them for each particular server.
Figure2. A traditional backup and restore model over a LAN with a primary server to control storage resources

Among traditional models, a more enterprise-oriented backup and restore approach alleviates some of the drawbacks of the dedicated tape approach (see Figure 2). This typically involves a primary backup and restore server that controls tape resources. Sophisticated applications, such as VERITAS NetBackup, Legato NetWorker, and Tivoli Storage Manager, control the backup and restore process. The backup server receives data from other servers across a LAN or WAN, then stores that data on centrally owned disk and tape resources. This centralized approach provides much better utilization of tape resources and makes the deployment of faster, more reliable tape drives and libraries much more cost-effective.

The primary drawback to this approach is that the network introduces potential bottlenecks to the backup and restore process that can impact the system's ability to meet backup and restore windows. In addition, using the primary LAN or WAN for backup and restore can degrade performance for production workloads running on the same network.

In contrast, Storage Area Networks (SANs) can accelerate and simplify the data backup and restore process. SANs are ideal for backup-intensive environments, especially when there are clearly defined areas for isolating backup workloads.

The switched 2 Gbit/sec full duplex capabilities of Fibre Channel fabrics can significantly improve backup and restore performance. Moreover, Fibre Channel is designed to transport large blocks of data with greater efficiency and reliability than IP-based networks. Two popular SAN-based backup and restore approaches are typically referred to as the "LAN-free" and "server-free" backup and restore models.

Figure3. A SAN-based LAN-free backup and restore model

Removing the LAN from the backup and restore process provides a variety of advantages. SAN-attached tape drives and libraries can be implemented so that each server sends its own backup data directly to the shared tape resources instead of through the network to the backup server. Sophisticated backup and restore software applications still control the process, tracking the backup and restore data. The SAN enables bulk data transfer from each server to shared SAN storage, but the LAN is used only for meta data communication traffic between the servers (see Figure 3). The result is a faster, more scalable, and more reliable backup and restore solution - with more effective utilization of storage, server, and LAN resources.
Figure 4. A SAN-based server-free backup and restore model

The SAN-based server-free backup and restore model is a newer approach. In this model, data is transferred directly between storage devices (for example, from disk to tape) without using host servers. This process is enabled by a technology called Third-Party Copy, which is implemented in SAN appliances (such as Crossroads or Pathlight bridges), host systems (such as Legato Celestra), or storage devices themselves (in the future). The server-free backup and restore model significantly reduces application host CPU cycles, thereby freeing up valuable CPU cycles to improve operating efficiency and enable higher workloads across the enterprise (see Figure 4).

The advantages of a LAN- or server-free backup and restore environment can be dramatic. For instance, after it deployed a LAN-free backup solution for its Windows NT Exchange server environment, a large North American hospital was able to reduce its backup window from 18 hours to just 4 to 6 hours. Other companies have experienced similar results.