By Heidi Biggar
For years, backup-and-recovery has been mostly about backup. What data should I back up? How can I make my backup process more reliable? Am I backing up everything that I should be, and in the allotted time? How can I make my backup environment more manageable?
While these types of questions still weigh heavily on most IT shops today, a larger challenge faced by administrators is not how to back up data but how to restore data quickly and reliably. And that increasingly involves disk, not just tape.
However, while the need for better ways to restore data are apparent, it's less certain whether users will actually spend money on new "rapid restore" technologies or whether they will rely on traditional backup-and-recovery applications or disk-array vendors for help with this growing problem.
"While a number of new technologies solve some issues, IT will still need to integrate them [with their existing backup strategies], and since users are looking to consolidate, not grow, their backup environments, this may be an issue," says Steve Kenniston, a senior analyst with the Enterprise Storage Group (ESG) consulting firm, in Milford, MA.
Another potential stumbling block for the new breed of products is user hesitance—especially among Fortune 1000 companies—to invest in unproven technologies. Nonetheless, Kenniston says that the need to keep increasing amounts of data online and readily accessible will challenge traditional backup/recovery paradigms as well as spur new forms of backup and recovery.
But if a recent study by TheInfoPro is any indication, users are looking seriously at rapid restore technologies. "Rapid restore capabilities" received a high "heat index" among the survey's 192 IT managers. The "heat index" took into account how soon respondents planned to implement the technology and the relative likelihood that the project would then be implemented.
Other backup-related projects that ranked highly in the survey include remote data mirroring, point-in-time copy, and automated backup.
Among those vendors betting on user interest in deploying rapid recovery products are Avamar, Data Domain, FilesX, Revivio, Storactive, StorageTek, TimeSpring, Vyant, XOsoft, and 3PAR. In some cases, the products are complementary (i.e., they can be used together to improve some aspect of backup/restore), and in other cases they are competitive.
While some of these vendors are currently shipping (e.g., Avamar, StorageTek, Vyant, XOsoft, and 3PAR), others will not do so for at least several months (e.g., Data Domain, Revivio, and TimeSpring).
Veritas, meanwhile, last month improved its recovery capabilities by releasing Veritas Edition for Microsoft Exchange, and 3PAR introduced Virtual Copy DBA for Oracle for use with its arrays.
Veritas' software is designed to automate and accelerate the recovery of Exchange 2000 servers. Support for SQL Server is expected later this month, with support for Exchange 2003 coming later this year. The software integrates with Backup Exec and NetBackup and supports any disk-based backup device.
"Our customers have been demanding this type of integration and with e-mail [first]," says Bob Maness, senior director of product marketing at Veritas. "Disk-based backup/recovery is part of our strategy for mission-critical applications—for both Windows and Linux."
Once the validity of the Exchange database is confirmed by the backup application, Veritas' software makes a point-in-time copy (or snapshot) of the database. The frequency of the snapshots is predetermined by the user and can vary from hourly to daily, up to a total of 31 images.
To recover lost or corrupted data, users simply click on the appropriate point-in-time image, and data is retrieved from the appropriate disk source. Additionally, administrators can run backups off the snapshot images, which improves application availability (since the application does not have to be taken offline to run a traditional backup) and speeds up the backup process.
3PAR recently integrated a snapshot capability into its storage arrays. Virtual Copy DBA for Oracle allows users to take hundreds of instant snapshots of an original volume. By using copy-on-write technology, 3PAR says it is able to minimize the amount of disk capacity used to store the virtual copies.
"Disk space is an issue with traditional snapshot applications," says David Scott, president and CEO of 3PAR. "Virtual Copy only consumes disk space for new writes, which means that users don't have to keep duplicate copies of data on-hand." Virtual Copy can be integrated with NetBackup for automatic archiving to tape.
So, what about the new rapid restore technologies? What are your options? What problems do they solve? And how do they integrate into existing backup/restore environments, or do they?
Some of the new products are stand-alone devices; others are software options. Some work at the file level, others at the block level. Some provide real-time data protection, others more-sporadic protection. Some incorporate snapshots, replication, and backup features into a single product; others just snapshots. Some are able to recover databases in minutes, others in seconds. Some work with existing backup applications; others replace them.
Whatever the case, all rapid restore products share a common goal: to improve some aspect of backup and restore. And that could mean speeding up recovery, automating and simplifying the backup/recovery process, reducing the amount of data that needs to be backed up and subsequently restored, or eliminating backup or risk windows—or all of the above.
ESG loosely breaks the new class of products into four categories (see table on p. 1): traditional snapshots (e.g., EMC and Network Appliance), next-generation snapshots (FilesX and StorageTek), intelligent "data capture" (Avamar, Data Domain, and Storactive), and real-time "data capture" (Revivio, Vyant, and XOsoft).
While snapshot technologies such as Network Appliance's SnapVault and SnapMirror speed up the recovery process and make the recovery more manageable, ESG's Kenniston says that there are some downsides to this approach, including disk space consumption (the backup data set is typically the same size as the original data set) and backup speed. "You gain availability, but you still have backup window issues," he says.
These types of products work with existing backup applications to make backups of the snapshots for disaster-recovery purposes.
Another approach is represented by products such as FilesX's Xpress Restore (see "FilesX focuses on fast data recovery," InfoStor, May 2003, p. 11) and StorageTek's EchoView. These technologies, explains Kenniston, capture snapshots at the block level (after an initial replica of the data is made) and move it into disk arrays. If a database is corrupted, the products are able to roll back in time to the last good snapshot and resume operation from there.
StorageTek's EchoView allows users to take frequent snapshots (every 5 to 10 minutes). These changes are journaled and saved to a separate area on disk. Users can keep a week's (or more) worth of snapshots (or views) online and they can back up to tape off the snapshots with standard backup applications.
EchoView works over iSCSI and currently supports Windows and Solaris servers; AIX, HP-UX, and Linux support are slated for this summer.
For users looking for continuous protection against system failures, there are several other options, including products from Revivio (not yet shipping), Vyant (RealTime), and XOsoft (Data Rewinder). These products capture data continuously, without using snapshots or replication.
"Revivio and Vyant not only make a copy of the data, but how it was created," explains Kenniston. The point is to be able to perform a function similar to an "undo" in Microsoft Word, which allows you to roll back databases second-by-second for real-time recovery, he explains.
"There is no moving or copying of data, which means you make 1.25 to 1.5 copies of your data, not multiple copies," says Kirby Wadsworth, Revivio's vice president of marketing. This translates into lower costs (because you're using less disk and tape) and faster recovery, he claims.
"It's all about being able to recover from any point in time, seamlessly, and without the use of snapshots," echoes Paul Parent, Vyant's president and CEO.
XOsoft takes a slightly different approach to continuous data protection. Instead of time-stamping data, its Data Rewinder product continuously captures application- and database-specific write and update events, and journals them for recovery purposes. If an error occurs, users simply "rewind" the I/O-operational event log back to the last consistent state—a process the company claims takes just seconds.
Data Rewinder supports Windows, SunOS, and Linux and is available for Exchange, SQL Server, Oracle, and Outlook databases; a server option is also available for recovery from file servers.
Vyant's RealTime works with leading database applications. Revivio will support most forms of data (e.g., databases and file systems) when it ships this fall. Data Rewinder and RealTime are pure software plays; Revivio's approach is based on an appliance that sits out of the data path and requires no agents.
TimeSpring is expected to enter this market in the second quarter of next year with continuous data-protection software that integrates backup, snapshot, and replication functionality.
Other players include vendors such as Avamar, Data Domain, and Storactive. While these vendors do not make products that facilitate the recovery process, they have tools that can significantly reduce the amount of data that is ultimately saved to disk, says Kenniston. "These technologies look at the data as it passes into their systems," he explains. "They check to see if the "blocks" live within their system. If they do, there is no need to write the block again." Kenniston says that these types of checks can potentially cut backup time as well as disk space.
Data Domain is expected to ship product in the second half of the year. For more information about Avamar and its "commonality factoring" capability, see "Disk-based backup options multiply," InfoStor, December 2002, pg.1.