Why does backup still matter?

Replication technologies give users access to more recent data, but replication in itself is not enough.

Recovery Point Objective (RPO) and Recovery Time Objective (RTO)


Figure 1. Translating RPO and RTO to technologies.

Backup is the only technology that provides:

• Protection against complete data loss
Traditionally, data protection covers 100% of data within a data centre, whereas replication covers much less, and to replicate all data would be too cost-, storage-, and network-prohibitive.


• Protection against viruses and data corruption
If replication is the only technology used, then viruses and data corruption will simply move from system to system. With backup you are able to go back to any point in time and retrieve a corruption- or virus-free set of data.


• Integrated archival management
Archival management is rapidly becoming a critical issue in most organisations, but by integrating it with backup there are obvious benefits, such as shared use of tape libraries, tracking and reporting of files, and much more.


• Integrated desktop/laptop protection (with DLO)
Protecting the data centre is only part of the equation, and with approximately 60% of all data stored on laptops and desktops (source: IDC), providing data protection across the entire enterprise is more critical than ever. Organisations need an integrated approach to protecting data, regardless of where it is.


• Cost-effective portability of data
Data is commonly stored on tapes, which are easy to move around from location to location. Compare the costs of moving a tape to the costs of moving a storage array and the cost-savings become clear instantly.


• Managed multiple point-in-time copies of data
With disk snapshots you can manage multiple point-in-time copies of data on disk, but sooner or later this will become cost-prohibitive and you will run out of storage. Backup, which is usually done daily, provides point-in-time copies of data going back in time much longer.

Backup, despite its many benefits, does not provide:


• Instant access to data (short recovery time objectives)
Restoring data from tape (or disk) takes time, anywhere from a couple of hours to days or even weeks, depending on how much data there is to be recovered. Even with all the data restored, the systems need to be setup and configured to run as they did before. Clustering, on the other hand, provides near-instant failover of applications from one site to another.


• Minimum data loss (short recovery point objectives )
For organisations requiring immediate access to data backup is not enough, even though backup is the cornerstone of a comprehensive disaster recovery strategy. Replication over IP or mirroring over fibre channel are the only technologies that will give immediate access to data at a secondary location following a complete site outage.

Backup and Disaster Recovery

Most organisations understand the necessity for disaster recovery solutions in today's economy, and as they become more dependent on data and uninterrupted access to data, the financial impact of a system or site outage quickly becomes obvious.

Potential Loss of Revenue by Industry Sector

Figure 2: Downtime can be costly, but more than revenue is at stake when outages occur

What is acceptable exposure to outage is reducing month by month. Some companies estimate the cost of downtime in 1000s of euros, while others experience a drop in stock price, loss of faith, customer satisfaction and loyalty, and so on. The higher the cost of an outage, the more likely that even a short-term outage can mean the difference between success and failure.

Traditionally disaster recovery planning looks at preventative action to avoid potential problems and look at how to deal with any unintended system interruptions; however, given today's business vulnerabilities, the requirement is quite simply to shorten the time to recovery.

Without the ability to restore data any disaster recovery planning, such as systems and network restore, is pointless. After all, it is the data that provides value to businesses, not the systems themselves. Organisations should set aside a considerable proportion of their disaster recovery budgets to make sure that data is backed up regularly and that backups are optimised for rapid restoration.

In spite of this, it is not unusual to find that organisations still do not backup regularly. Some do backups but do not check or test to see how effective these backups are. In one case an organisation had been religiously backing up every day, but when the disaster occurred they turned to their off-site backup tapes to restore the data to the systems and found the tapes blank. No one had tested the system to see if anything was actually being backed up on the tapes. Making mistakes like this can be very costly, not only in terms of lost data if a disaster occurs, but consider all the tapes used over the years to backup without anything on them.

Other organisations have tested for data integrity, but not tested their ability to restore the data backed up. In fact, it would not be surprising to find that most organisations have not changed or upgraded the backup of this fundamental business asset for several years. The process of backing up mission-critical data to tape and then taking those tapes off-site (most disasters occur locally) on a scheduled basis has the advantage of dispersing the data asset; however, the downside is that time to recovery can be slow.

Technologies to Protect Data

In today's environment we have several alternatives to these traditional data protection methods. One of those is vaulting. The VERITAS NetBackup Vault option simplifies the process of rotation and retention scheduling for off-site copies, by automating the process.

The Vault option fully integrates with the NetBackup job scheduler and provides a Graphical User Interface for configuration flexibility. In essence, the remote tape is treated as a locally installed system by connecting the tape system to the host system. The transfer of backed up data is completed with no negative impact on performance, regardless of the distance involved.


Figure 3: The NetBackup Vault Option simplifies management of off-site tapes


Figure 4: Re-using tapes after expiry is easy and lowers media cost

Another option is remote mirroring and replication. VERITAS Volume Manager and Volume Replicator allow the creation of duplicate volumes of data on a physically separate storage system. This duplicate set of data constitutes a real-time data backup. The advantage is that the duplicated data is restored as if it were read directly from a set of disks rather than having to be copied and reloaded to disk from the tape backups.

Business can then continue in a matter of minutes rather than hours or even days. The down side is that if the data mirrored is corrupt in any way then the mirrored version is also corrupt. To cover this eventuality traditional backup methods are essential. (The next edition of the Disaster Recovery Newsletter will cover Replication and Mirroring in-depth)

Tape vaulting and remote disk mirroring are not a total solution, but they are certainly technologies that should be incorporated into a comprehensive disaster recovery plan. Ultimately, companies will find that a mixture of traditional vaulting, remote mirroring and replication is applicable to their business depending on the criticality of their particular systems and applications. Not all data and systems are created equal, so each organisation must decide which technology is right for its need.

New Data Protection Technologies for Disaster Recovery

New technologies in data protection can also help in a disaster recovery scenario. Disk staging, for example, can reduce backup and restore times by having the data stored on a disk, which is traditionally faster than tape.

The downside of disk staging is that if the disaster is site-wide then the data on the disk is likely to be lost as well. Synthetic backups is another new technology that can help in a disaster scenario by only backing up a fraction of the data using incremental backups, the amount of data to be restored can be reduced considerably, thereby getting back to business faster.

Bare Metal Restore is also a fairly new concept, whereby systems are restored in a matter of minutes rather than hours or days. Bare Metal Restore creates images of systems that can then be restored in minutes should a disaster occur and once the system has been restored the recovery process from tape or disk will start automatically. This automatically shortens the restore time and automates the process of re-building servers.

In one example a customer had a server crash, and it took 48 hours for them to rebuild the server as they had to first identify the problem, install the operating system, install the applications, apply all the patches, setup the network configurations, and make sure that everything was running. The hard part was finding all the right installation disks and configuration settings, which is why the recovery was so slow.

Even after the server had been rebuilt the restore from tape took a long time. So all in all, the entire server was down for several days. With a Bare Metal Restore solution, the 48 hours of server downtime could have been reduced to less than 10 minutes, which is what it usually takes to rebuild a server (these figures have been verified by several customers).

Conclusion

Given the explosion of data and the reliance on IT systems for generating revenue, any kind of downtime carries with it a significant cost, both tangible and intangible. Disaster recovery planners should therefore carefully consider their requirements before looking at what technologies are available.

The best plan is one that builds on backup as the cornerstone of a solid disaster recovery strategy and then adds on more advanced technologies to meet higher service levels. By employing a mix of different technologies organisations can ultimately achieve 24x7 business continuity, maximizing productivity and usage of systems for generating revenue.