Why Use Replication?

It used to be that running a daily backup was the only protection required, but as internal employees, partners and customers exceedingly rely on access to these systems around the clock, providing higher levels of protection becomes vital.

Replication is the only technology that can satisfy the needs of the most demanding systems, as only replication can provide instant access to data and zero data loss (minimal Recovery Point Objective or RPO and minimal Recovery Time Objective or RTO).

Replication is simply the process of copying data from one location to another, but there are many ways in which this can be done and this article will explore some of these ways.


Figure 1: Data loss (Recovery Point Objective) and Downtime (Recovery Time Objective)

Backup is still the cornerstone of a solid disaster recovery strategy, but since backups are usually run only once a day there is a high risk of losing large amounts of data and also having systems being offline for long periods of time (the Recovery Point Objective and Recovery Time Objective for backup is usually days, whereas with replication it is usually minutes or seconds).

With replication, none of these risks exist, but a comprehensive disaster recovery strategy should not only use replication as the sole means of protecting data as there are some disadvantages of doing so. For example, if a virus enters the system and only replication is used to protect data then that virus will most likely be replicated along with all the rest of the data to the secondary location.

That is why backup should still be used as it is a point-in-time copy of data on tape in a safe location that can be restored without viruses or data corruption.

Using replication (especially in real-time) for 100% of data in an environment is unrealistic and cost-prohibitive. The bandwidth required for most businesses wanting to replicate all data would be astronomical, besides the fact that all data is not created equal so it would not make sense to replicate everything. Rather consider these kinds of systems:


• Customer-centric systems: All data stored in systems being accessed by customers is mission-critical and should be protected at all costs.


• Partner-centric systems: Most of the data stored in partner-centric systems is business-critical, but not all. Carefully consider what should be replicated and what should only be backed up.


• Internal systems: Certain internal systems are critical for keeping the business going, such as e-mail, and should be protected at all costs. Other functions might accept a day or more worth of data loss.

Different Ways of Replicating
When it comes to replicating data there are many ways in which this can be done and many technologies used to do so. When choosing the right method of replication, carefully consider the data that is being protected and how much of that data you can afford to lose. The different methods are:


• Synchronous Replication: Data is copied in real-time from system to system, so if the primary system was to fail the secondary system has an exact copy of data and can take over instantaneously. Synchronous replication is the only method that guarantees zero data loss.


• Asynchronous Replication: Data is copied from system to system with a small lag time, usually measured in milliseconds. This form of replication is less dependent on available bandwidth than synchronous replication, but if the primary system was to fail some data loss would occur.


• Periodic Replication: Data is copied from system to system on a set schedule, for example once a day or twice a day. Periodic replication is not ideal for critical systems as large amounts of data loss would occur if the primary system was to fail. Periodic replication is ideal for remote office backup.

Once the criticality of data has been identified, then a suitable method of replication can be selected. After selecting a method of replication, comes the selection of what technology to use. The two most common ways of replicating data are server-based replication and storage array-based replication.

Server-based replication has the advantage of not locking users into using a certain storage array from a certain vendor. The replication solution from VERITAS for example, can replicate from any array to any array regardless of vendor, such as from EMC to Hitachi, or IBM to Serial ATA disk.

This ultimately lowers costs and provides users with the flexibility to choose what is right for their environment. Most server-based replication solutions can also replicate data over IP networks natively, so users do not have to buy expensive extra hardware to achieve this functionality.

Storage array-based replication has the advantage of being operating system independent, since replication is done from array to array. The downside of storage array-based replication is that vendors often demand users only replicate from and to similar arrays.

This can be very costly, especially if you are using high-performance disk at your primary site and now have to use the same at your secondary site. Also, array-based solutions often require extra hardware to send data over IP networks, further increasing costs.

For a full explanation of server and storage array-based replication technologies and their advantages and disadvantages, download the analyst article VERITAS's Diverse Server-Based Replication Technologies Meet Multiple Business Continuity Needs by DH Brown in this editions download section.

Distance Matters
When it comes to choosing the right technology and method of replication, another factor has to be considered: distance. Most businesses use some kind of volume management technology to manage their storage arrays and disk systems, and can therefore replicate data over fibre-channel environments natively (often called mirroring). There are, however, some limitations to replicating over fibre-channel, such as distance and method used.

The distance between two systems when using fibre-channel replication needs to be less than 100KM or else write latency becomes an issue. This has been verified by numerous studies (for more information download the article VERITAS Performance Brief - Remote Mirroring - Using VxVM for Metropolitan Disaster Recovery in this editions download section).

The other downside, besides distance, is that you can only replicate data synchronously, which in most cases is not a problem, but could potentially overburden the network. For distances larger than 100KM, replication over IP networks is the only technology available today.


Figure 2: Choosing the right replication technology depends on the distance between sites


IP-based replication has the advantage of being able to replicate data synchronously or asynchronously over large distances. There are several reasons for choosing either ways of replication, but the main ones are:

Synchronous Replication:

• Advantage: Zero data loss, as data will not be written on the primary system until it has been written on the secondary system. This is the highest level of real-time protection available.


• Disadvantage: Potential to impact performance of applications using replication to protect data, since data-writes are not acknowledged until the write has been written by the secondary system. Only a real issue if distance is too great or bandwidth connectivity is unreliable.

Asynchronous Replication:

• Advantage: Zero impact to performance of applications using replication to protect data, since data-writes do not have to go across the wire to the secondary system to be acknowledged on the primary system. Ideal for extremely long distances, such as from London to Sydney.


• Disadvantage: If the primary system was to fail there would be some data loss, which is usually measured in milliseconds. For some systems, milliseconds of data loss is acceptable, but for the most mission critical systems even 1 lost transaction can mean millions in lost revenue.

Using Replication to Protect Remote Offices
Protecting remote offices can be very costly, especially since having IT Administrators located in each location quickly adds to the overall costs of managing remote data. Plus, each system needs to have a tape library attached to it so that data can be backed up daily.

Using periodic replication to protect remote offices can lower the cost of administration and hardware, by simply replicating data on a set schedule to a primary site where everything then gets backed up. Using VERITAS' solution, this is easily done with the administrator selecting which systems in the remote offices to protect and then creating a schedule on when to replicate data.

Even more, the administrator can pick and choose what gets replicated and how much bandwidth is used for the replication so it doesn't affect the employees in the remote offices. The amount of data that gets replicated is equal to what changes there are in the remote offices.

For more information on protecting remote offices, download the analyst article VERITAS Storage Replicator: A Step in the Right Direction Blending Backup and Replication by IDC in this editions download section.

Summary
A comprehensive disaster recovery strategy should not just use backup, but enhance protection of mission-critical systems with replication so data loss can minimized.

Picking and choosing the right method of replication is not always easy, but by mapping out what applications you have it will become easier to figure out what needs to be protected at the highest levels using replication.

The real question becomes, 'can you afford not to protect your data?' as any kind of data loss carries with it severe consequences.