Recovery management: focusing on recovery-oriented data protection.

Introduction

If the whole purpose of backing up data is to support recovery, why do products in this space focus on optimizing backup operations instead of recovery operations? Many backup product optimizations in the last 10 years that have made backup easier have in fact complicated recovery operations. That doesn't help when recovery operations are required. What is needed among vendors is a shift in focus from optimizing backup to what enterprise users really need: a better way to optimize application-level recovery.

Defining Recovery

At first glance, many administrators associate "recovery" with major disruptions, events like a server or site failure. But recovery can also include much smaller events like an inadvertently deleted file or table or logical corruption introduced into a data service (file system or database) by some form of user error. In either case, administrators are required to go to an image of the data at a previous point in time and perform recovery operations, but there is a huge difference in scope. There is also a large difference in frequency: recovery requests to address small failures like dropped files or tables happen much more frequently, often several times a week or more. The key point in defining recovery, though, is that recovery is not just about moving blocks of data, its about recovering applications (or subsets of those applications).

[FIGURE 1 OMITTED]

Solution Characteristics

Despite the advancements in backup, it still has issues. On the data capture side, the biggest of these is probably the nature of the impact on production applications; service disruption and/or performance degradation typically occur during backup operations. But the focus of a better solution cannot stop there--it must also address recovery point objective (RPO) and recovery time objective (RTO) issues on the recovery side. Enabling faster backups and restores through the use of disk media can be part of the solution, but it is not the whole solution. To provide a solution, the architect must focus on the key requirement: enabling faster, easier recovery of the affected application service with a minimum of impact to production operations.

Current data protection operations offer limited recovery points--there's rarely more than one backup taken per day and hardware snapshots may only be taken several times a day at most due to storage and cost considerations. These methods require an administrator to pick the recovery points (perform the backup or snapshot) before they know what type of failure they must recover from (physical, logical) and when it will occur. The available recovery points created using these conventional methods rarely if ever results in an optimal recovery point for any given type of failure. A more desirable option would be to track and present all possible recovery points and allow the appropriate recovery point to be selected after the failure has occurred. There is a technology available today called "continuous data protection" (CDP) that enables this and it can provide an excellent technology foundation to a recovery-oriented solution.

A valuable characteristic of CDP is that it removes the need to perform "incrementals" of any type, either on the backup or restore side. Data capture is continuous on the production server side, so the concept of "incrementals" just does not apply. On the restore side, once a recovery point is chosen, the associated snapshot is created and presented in a single operation--there is no need to incrementally apply anything to get to that point. This results in a much simpler, much faster recovery.

But CDP by itself is not enough. In using CDP, you will have the optimal recovery point available for any failure type, but CDP alone does nothing to help you select what that point is. What is needed is additional intelligence about application states at various points along the CDP timeline. By correlating the timeline with information about application processes, such as the completion of a database checkpoint, pre- and post-patch points, quarterly closes, etc. this solution could provide the administrator with the ability to select the optimal recovery point for any given type of failure after it has occurred.

CDP technologies support a data capture component that addresses backup window and RPO issues relative to existing data protection products, but the nature of the data capture component--sometimes called an interceptor--has important implications in enterprise environments. Applications that need protection may be hosted in different locations throughout a network computing environment. If the data capture component is lightweight and completely independent--not tied to a particular operating system, volume manager, file system or application--it can potentially be located across a variety of network device types (servers, switches, arrays, blades, etc.) and operating environments. In this way a single recovery engine type can be used to protect applications regardless of the environment in which they are hosted.

Operational considerations also mean that any new recovery management solution can be much more easily adopted if two things are true:

* It complements and enhances the existing data protection infrastructure

* It is available from trusted enterprise storage suppliers

Customers have a large investment in existing data protection infrastructure and while this infrastructure may be suboptimal in terms of backup window, RPO and RTO, it does a very good job of meeting archival requirements. A "rip and replace" strategy is just not viable in this area. The recovery management solution can "front end" the existing infrastructure, resolving backup window, RPO and RTO issues for near term recoveries (on the order of days), but it must also enable policy-based data staging to enterprise backup software and its associated media. In this way, existing data protection processes are only slightly modified and companies can stay with their current processes for all archival and off-site data storage. The only time an administrator would use a modified recovery process is when there is an optimized recovery option available.

The Importance Of Trusted Suppliers

Innovative storage technologies can be interesting, but if they are going to be widely deployed enterprises must depend on obtaining the solutions from their trusted enterprise storage suppliers. Recovery management is an infrastructure layer; (Figure 1) and for service and support reasons enterprises are less likely to take a chance on a startup player's products in this space regardless of how innovative it may be. The availability of that same solution through a trusted supplier can provide the peace of mind of knowing that the vendor will be there to help install, configure, support and enhance the solution on a worldwide basis for years to come.

Candidate Applications

Recovery management can be used to overcome the limitations of existing "point in time" oriented data protection products for application environments where backup window, RPO and RTO are issues. The types of applications most likely to benefit from recovery management's transparent data capture and optimized recovery point and time capabilities will have the following characteristics:

* The application runs continuously (7X24 operations)

* High rate of data change (discrete snapshots "age" very quickly)

* High impact to company of application downtime agement can be thought of as a technology that will allow enterprises to meet very strict service level agreement requirements with minimal or no impact to production environments. This will tend to be a company's mission-critical applications, regardless of whether they are running on structured (database) or unstructured (file system) data services.

Conclusion

Storage analysts agree that enterprise data protection is an area that is ripe for innovation. CDP is an interesting foundation technology that Data Mobility Group, Enterprise Storage Group and Taneja Group have all commented on publicly and they are all predicting that it will make its way into enterprise storage infrastructure over the course of the next several years.

By combining time addressable storage (CDP), a technology that addresses backup window issues and supports an ability to recover to any previous point in time, with event addressable storage, a technology that enables a view into the state of application processes at any point along a timeline, vendors can offer an application recovery-oriented solution that offers optimized RPO and RTO characteristics for any type of failure. An ability to integrate seamlessly with existing data protection infrastructure and availability of these recovery management solutions from trusted enterprise storage suppliers will help overcome obstacles to adoption in production enterprise environments. Recovery management reflects a change in focus from backup-oriented solutions to recovery-oriented solutions--a change that addresses the key customer requirement of enabling faster, easier recovery of affected application services with a minimum of impact to production operations.

Eric Burgener is vice president of marketing at Mendocino Software (Fremont, CA).

www.mendocinosoft.com