Breakthroughs In Enterprise Backup Solutions For NAS File Servers.

This article is one in an ongoing series related to the Open Storage Networking (OSN) initiative.

Data from Gartner Group validates that NAS has the lowest total cost of ownership (TCO) of any storage architecture: storage area networks (SANs) are nearly 10% more, external RAID is 70% more and remote storage service providers (SSPs) are 140% more. NAS appliances are now becoming the de facto building blocks for a wide range of storage solutions--from workgroup and department "home" file servers through high-availability enterprise-class storage servers.

So, what's the problem? Organizations have found their easy-to-deploy, easy-to-use filers not so easy to back up. This creates a new challenge: provide data protection for the hundreds of gigabytes, and often terabytes, of data residing on these filers--data that demands protection.

Fortunately, IT managers can now select from an equally wide range of backup solutions to protect their NAS filer data. There are quick and simple backup appliances appropriate for backing up workgroup and department filers. There are also new enterprise-class backup solutions that offer all the performance advantages touted in storage area network (SAN) backup solutions. The most recent high-end backup solution even anticipates one of the upcoming technology trends in the rest of the storage world and backs data up over Gigabit Ethernet (GbE).

Of the many solutions used to back up and restore data on NAS filers, this article compares and contrasts five popular solutions. The simplest of these backs up data over the local area network (LAN) from the filer to a backup server. The other four solutions utilize the Network Data Management Protocol (NDMP). These are the local, filer to filer, filer to Fibre Channel, and filer to GbE backup solutions.

Backup Over The local Area Network

In this solution, which is shown in Figure 1, the backup server "mounts" the various NAS filers that are on the LAN and backs up the filer data across the LAN. The data comes across the LAN in either Network File System (NFS) or Common Internet File System (CIFS) format. The backup software and the file system on the server convert the filer data to block format data, which is then written to the tape drives in an automated tape library.

This solution is suitable for backing up NAS devices at the workgroup or department level. It is simple and easy to implement. Furthermore, if the NAS device does not support NDMP, this is the only practical way of backing up the data.

There are a few limitations to this solution. One limitation with this approach is that "meta-data," such as security attributes, access control lists, etc., can be "lost" with this type of backup approach. Some backup packages have manual work arounds for backing up meta-data not accessible through the file system interface. Also, some NAS vendors have addressed this issue of preserving the meta-data working with backup software vendors to port agents to the NAS operating system. The availability of either of these solutions varies across NAS platforms and backup software packages.

Another limitation with this solution is that backing up more filers or larger capacity filers might congest the client LAN, thereby reducing its performance to unacceptable levels. The use of LAN switches helps to mitigate this congestion by minimizing the impact to clients not using the server. However, this does nothing to minimize the impact to clients trying to use the server that is being backed up.

It is also possible to mitigate this congestion by offloading the backup traffic onto a separate backup network. This can be accomplished if the NAS devices support a second Ethernet port that can be used for the backup data.

In this solution, the backup server itself could be dedicated to the task of backing up data across the local area network. Therefore, it lends itself very well to being a "backup appliance." Similar to a NAS primary storage appliance, a backup appliance would be easy to install, easy to use, easy to manage, and, because it is dedicated to one function, would be highly reliable. One example of a backup appliance that could be applied to this solution is Quantum/ATL's LANvault product.

Filer-Dedicated Tape Backup

The remaining four solutions all utilize the open NDMP standard (www.ndmp.org). NDMP is a client/server standard that separates the I/O-intensive data path from the control path. An NDMP "client" resides on a backup workstation that runs a popular standard operating system, such as Windows NT, Windows 2000, or Solaris. This client performs all the administrative control functions, such as backup scheduling, keeping track of the backup database, etc. The NDMP client also initiates and controls the backup operation by communicating with the NDMP servers that reside in the NDMP-compliant NAS filers. Note that, unlike in the previous non-NDMP solution, the backup workstation--the NDMP client--does not process the I/O intensive data; the NAS filers process all the backup data in response to the commands from the NDMP client.

This separation of control and data paths is commercially very important to the ongoing development of NAS backup solutions because there is a clear and open interface between the NDMP client and the server. The independent software vendors (ISVs) can focus on developing the administrative control functions associated with the NDMP client on standard operating systems. Each NAS supplier, in turn, focuses on developing the data I/O-intensive NDMP server functionality within the NAS device.

In the filer-dedicated tape backup solution, which is illustrated in Figure 2, the NDMP server functionality is embedded wholly within each of the NAS filers. The NDMP client kicks off the backup operation and keeps track of all the files that are backed up onto the tape drives. In response to the NDMP client, the NDMP server reads the data from the NAS filer, converts the data to block format, and sends the data across the SCSI connection to a standalone tape drive or a tape drive within an automated tape library.

This solution is most appropriate when there is a single large filer on the. network. The key advantages of this solution over the first solution is that it provides full support for the meta-data attributes, no backup data flows across the LAN, and the backup performance is high. The disadvantage is that the filers do not share the backup resources, such as the tape drives and the automated tape library.

The problem with scaling this solution to large numbers of filers is that it increases the effort required to manage all the separate backup devices.

Steve Morihiro is the CTO and vice president of engineering and Rory Bolt is the director of systems design at Quantum/ATL (Irvine, CA).

This article is the first in a two-part series. The second part will appear in the April issue of CTR.