How Snapshot Technology Will Change The future Of Backup And Recovery.

Providing a new level of application availability

Prior to the Internet revolution of the early to mid-1990s, backup and recovery was a fairly mundane task that occurred solely at the end of day. Most system or database administrators would shut down systems at night and perform either an incremental- or full-backup to make sure data was fully protected. While the operation was slow because of a variety of factors (i.e., slow networks, slow hardware, simplistic software, etc.), backup windows were quite large and jobs easily completed within this timeframe. Restores were also slow, but since data was captured easily in these large backup windows, performance was not seen as that crucial as long as the data could be recovered. A variety of architectures were also used for these operations (Figure 1).

As time progressed and corporations expanded globally, the idea of shutting down systems for backups became unacceptable. Soon backups needed to be taken while systems were online and this process took its toll on the end-user population. Suddenly network collisions were becoming more and more frequent as backups generally took a larger percentage of network bandwidth. At the same time, performance suffered for end users that continually tried to access applications. This would not only aggravate a corporate workforce, but also affect the bottom line.

Today the situation has only gotten worse as data growth in enterprise IT environments continues to accelerate between 100 and 400 percent per year, while data-protection solutions have struggled to keep pace. Backup performance is crippled by the needs of online all-the-time e-business applications, and backup windows are no longer present. Restore performance continues to be slowed by the inability of many software products to retrieve data quickly and efficiently. Solutions for these environments need to look at alternatives to conventional file-by-file backup methodologies.

Enter Snapshot Technology

While snapshot technology has been around since the early to mid-1980s on the mainframe, its use in the open-systems backup and recovery space is a relatively new phenomenon. The simple idea is to make a point-in-time copy of data that can be backed up without impacting systems, like a traditional backup, while at the same time offering the ability to restore data when it is needed. To put it another way, the concept of snapshot-ing is taking a stable image of data that could be changing. The goal of any snapshot backup and recovery solution is to provide a new level of availability for applications that can never be offline or experience business interruption.

From an operations point of view, the steps that a system goes through to perform a snapshot based backup are as follows:

1. Backups are triggered by one system sending a command to another system saying that a backup is about to occur.

2. When this signal is received by the target machine, a quiesce of the application takes place.

3. Next, the file system is flushed so that all pending transactions are completed.

4. From this point, the snapshot is generated.

5. After the snapshot, the application is released and can resume its normal operations.

6. The snapshot is then used to create a map that translates the logical name of the data into its physical location on disk.

7. From this point, the map is used as the data to be backed up to either disk or tape.

Types Of Snapshots

While processes can vary significantly from the steps listed above, the underlying goal of creating a stable image of data is the same no matter which snapshot methodology system or database administrators use in their environments. Snapshots can take many forms when it comes to data protection and each approach has its strengths and weaknesses.

* Hardware Snapshot Technologies

* Strength: No CPU overhead during the snapshot process

* Weakness: Volume level snapshots; More disk required

Hardware snapshots, often known as mirrors, are copies of data that usually reside on disk arrays or subsystems. In order to back up this data, the mirror is split and can be backed up without affecting production systems. After the backup is complete, the mirror is resynchronized and is once again a complete data image. Although this snapshot process does create an exact copy of data, it requires additional disk space and needs to restore an entire volume rather than an individual file. This type of backup also requires a software backup and recovery solution that integrates with the hardware device. This is necessary because of the fact that time stamps are applied to these snapshots or split mirrors. The software must be able to recognize these time stamps so that when a restore operation takes place, data is correctly organized and data corruption is avoided.

* Software Snapshot Technologies

* Strength: File level granularity 

* Weakness: CPU utilization on the host server; Lack of support from application and database vendors

Most software snapshots are implemented through a copy on-write technique that takes a picture of a file system but also allows updates during the entire process. The process of taking a snapshot is almost instantaneous. Once this happens, a data map is created from the snapshot and this data is moved to a backup medium at the block level. Any block that is changing while this image is being backed up is moved to cache. It is important that the file system remain in production while also enabling the snapshot to preserve the data at that exact moment in time. When the backup comes to the point where a block was updated, it simply goes to the cache and reads the original block. In the event of a restore, data can be restored more efficiently at the file level instead of the volume level. This is made possible by the map of data that is preserved during the backup.

Other software level snapshots can be focused around applications or databases. If a particular application or database vendor supports snapshot technology through their defined backup and recovery APIs, snapshot based backup and recovery becomes a viable solution. Unfortunately, only a few application and database vendors support this kind of technology today, making it impossible for third party solutions to provide this type of functionality.

* Storage Area Network (SAN) Snapshot Technologies

* Strength: Resource and Cost of Ownership savings

* Weakness: Lack of SAN and third party copy technology acceptance

The latest trend in backup and recovery technology, and a theme likely to evolve in the near future, is to combine snapshot technology with new data movement techniques within a SAN environment. While the early stages of SAN based backups have been extraordinarily successful in moving traffic to a devoted network, snapshot technology can expand SAN capabilities even further. SAN snapshots can also incorporate third-party copy technology or facilitate what is known as "server-free" backups. So now, not only are resources taken off the LAN and devoted to the SAN, but the server impact in the backup can also be reduced. With this new technology a server or intelligent SAN device (Figure 2) can move data directly from disk to tape, avoiding the application or database server altogether. In some cases it may be desirable to send data through at least one server dedicated to backups. While this may not seem like true third party copy or "server-free" backup, it really is due to the fact that only one server is inv olved in the data movement process. Also, this architecture will allow users to apply techniques like multiplexing to improve backup performance. However, in order to moved at a without an application or data-base server, data needs to be frozen at a point in time. In most cases, this snapshot occurs at the data host, and then a hardware device that supports the SCSI Extended Copy Command. (a block level command that makes device-to-device data movement possible) does the actual data movement. Users can now not only reduce impact on production systems, but also leverage investments made in SAN technology by moving the entire backup data movement process to its own dedicated network.

The Future

Eventually snapshot technology improvements will create IT environments where backup windows are irrelevant and recovery times become a fraction of what they are today. This most likely will take the form of integrated snapshot techniques at the host or subsystem level or even the ability to have a server-free restore of data. The possibility also exists to incorporate many of these technologies together in other ways so that users have a variety of options when it comes to implementing this kind of solution.

In addition, snapshot solutions will continue to provide the following benefits to users both today and in the future:

Zero downtime backups. Backups can be run during business operations while producing zero impact on the host server. This eliminates backup windows and keeps businesses running 24x7x365.

Improves restore performance. By providing more efficient backup techniques, users are to restore more up-to-date data and also eliminate delays.

Server+Free Backups. Snapshot technology paves the way for more advanced solutions such as server-free backups, which eliminate backup overhead in the application server (CPU and I/O Overhead).

Leverage SAN and Application Investments. Snap-shot technology, used in a SAN or with a particular application, will enable more functionality than ever before when it comes to backup and recovery.

Michael Adams is the product marketing manager at VERITAS Software (Mountain View, CA).