This article can also be found in the Premium Editorial Download "Storage magazine: Upgrade path bumpy for major backup app."
Download it now to read this article plus other related content.
|Three types of CDP|
Block-based Continuous data protection (CDP) excels at transparent data capture and presenting views of different points in time. It can require additional integration work with the application. Some CDP products support a tag to specific "times" that are matched with application events, such as quiescing a database, to allow for discrete recovery points.
Application-based CDP is specific to a particular app, such as Oracle Corp.'s 10g. It's responsible for performing all of the continuous protection and journaling necessary to roll back to any point in time. Queries, rows, columns, tables, transactions or the entire database can be rewound to any point in time without disrupting the running application. Neither block- nor file-based CDP usually has that level of visibility. The value of application-based CDP is its extensive application-awareness.
File-based CDP products typically run on the protected app's servers or workstations. Most are agent-based and are conceptually similar to application-based CDP but, for these, the app is the file system. A key advantage of file-based CDP is its flexibility in setting policies for different file groups. Recovery point objectives and recovery time objectives can vary widely based on each group. Block-based CDP doesn't have that level of granularity. Recovery is also more granular because it can be to any point in time on a specific file or group of files.
The many faces of CDP
A CDP product can be file-, block- or application-based (see "Three types of CDP," at right). CDP may be provided as a standalone product; as a feature of backup, replication or database management software; as a separate backup appliance; or as array-based data protection software. Some CDP apps are agent-based, while others don't use agents at all. A CDP product may run on a LAN appliance or on an intelligent switch (see "A sampling of CDP products," below right).
And some products that claim to be CDP don't really fit into the category because they lack either a continuous protection capability or any-point-in-time recoverability. For example, Microsoft Corp.'s Data Protection Manager (DPM) lacks continuous protection and the ability to recover from any point in time. Numerous server replication products capture every change continuously but can't rewind beyond the most current replication. That type of replication is analogous to mirroring and protects only against hardware failures.
CDP and Microsoft Exchange
CDP is designed to provide the highest level of data protection for apps that can't afford to lose any data, such as database management systems, point-of-sale systems, financial transaction systems and e-mail.
The main app driving CDP sales these days is Exchange, which is extraordinarily difficult to restore with most data protection apps. Restoring Exchange is a complex and frustrating endeavor, and can be particularly time consuming depending on the number of mailboxes and messages that need to be restored. Typically, these are the major steps of an Exchange restore:
- Apply the last full backup.
- Apply the transaction logs (if they're available).
- Restore messages and transactions to each mailbox; this is time consuming, so this step is often skipped, leaving a lot of data that's never restored.
- During the restoration process, Exchange is down and a temporary server is required.
|A sampling of CDP products|
|Click here for a comprehensive list of CDP products (PDF).|
The benefits of CDP for apps other than Exchange are less certain. For example, most databases include an older, well-known variation of CDP called journaling. In a disaster, a DBA can restore the database from a known good snapshot or backup and then journal forward to the transactions that occurred after the good copy was recorded to recreate an up-to-date restored database. Oracle Corp. has gone beyond simple journaling by incorporating CDP into its latest 10g database. A DBA can rewind the database to a consistent version at any point in time with the click of a mouse and then journal forward to all of the transactions that have occurred since the time the version was created.
This was first published in September 2006