The scramble is on to solve recovery puzzle: virtual tape, MAID, SRM are some leading solutions

Computer Technology Review, Sept, 2004 by Mark Ferelli

If you took a poll, the most popular buzzword in the mass storage industry would be "information lifecycle management." But with all the developments that we have seen in storage--disk-to-disk backup, disk-to-disk-to-tape backup and archival, storage resource management software, virtualization solutions, better connectivity--the issue of recovery continues be a primary concern in the data center. This is true for the enterprise, the small-to-medium enterprise, and the increasingly popular small-to-medium business segment of potential storage consumers.

Regular readers already know that the management of storage is the management of complexity. The management of recovery is absolutely the same. As veteran Gartner analyst Carolyn di Cenzo says: "Recovery is getting more complex as more options are made available. It's hard to keep track of what the best recovery vehicle is if you take advantage of all these new things."

From Replica to Virtual Tape

Areas that seem to bear watching are the replication and virtual tape spaces. Historically, replicas are volume-level recovery tools, and in a disaster recovery scenario this kind of replication is important. But it is a waste of time and effort to restore a single file from a volume. The trick is to provide volume-level replication with file-level recovery. Di Cenzo points at the work that Network Appliance and Sybase have been doing. Sybase, especially, offers their V21 Protector, which is a bare metal recovery product. It is possible to create the bare metal images while the server is live--no shutdown. The image can be updated incrementally. It also does single file recovery.

Virtual tape libraries (VTLs) also offer a popular alternative with such companies as IBM, Quantum, Overland Storage, and FalconStor participating. With virtual tape solutions, data is backed up on a nearline disk array (usually SATA disks) and then later moved to tape. Nearline storage offers higher transfer rates, but typically not with the economy of a tape solution. One of the questions that an end user might ask an integrator is whether the VTL solution will support all the agents that are required to do a complete server backup. Di Cenzo brings up the example of Avamar, which does file backup, but doesn't have agents for Exchange or Oracle.

Perhaps the most appealing thing that VTL brings to the data center is familiarity. Backup to tape is a known quantity, with years of use in almost every installation. A user gets the disk speed advantages, but there is no need to reinvent backup for the data center. Recovery may be from disk, but the operation has the look and feel of mounting a tape. They can clone to a real tape for offsite storage. If the user clones the data, the backup software remembers that you have the clone--an advantage of media management.

Speed is often of the essence, especially with the backup window shrinking to extinction. Companies such as Arkivio address this crucial problem.

One newer approach addressing the issue comes from Massachusetts-based Sepaton. This company offers a synthetic full backup by dramatically reducing the time it takes to perform a full backup operation. Unlike other approaches, this application runs on the SEPATON S2100 data protection platform and integrates with currently deployed backup/restore infrastructure. By leveraging the S2100's ContentAware architecture, synthetic full backup builds full-backup datasets in minutes versus days by removing the time-consuming step of physically copying data to build the new backup. This breakthrough allows end users to have off-site replications of their data days before it would be available using traditional processes.

Another high-powered disk appliance leveraging virtual tape comes from Copan, who has partnered with FalconStor software. The value in this appliance is that it is partitioned and managed in such wise that it is easier to use in a backup application.

Copan uses a MAID-based technology that allows Serial ATA (SATA) disks in an array to be started and stopped when access to a particular piece of data, on a specific drive, is needed. Copan thinks it can deliver a fast, reliable disk-based solution for backup and recovery at a tape-competitive cost. Their array architecture is designed to prolong the life of SATA drives, which were never designed to spin continuously, at SCSI duty cycles.

SRM Needs Cross-Platform Performance

Storage resource management (SRM) software is often looked at as a solution for the problems of recovery. This makes sense when you consider that the resources that lend themselves to fast recovery are evaporating. Business executives that know more about distressing budgets than storage management are telling administrators that they will have to manage more terabytes with the same number of people. They have not yet realized that "doing more with less" is a myth that can only cause less efficiency in the data center rather than more.

To even flirt with a solution requires greater abilities to manage automation. And that is a vision of SRM. But until recently, effective integration of backup/restore with resource management has been anything but on the fast track. Anders Lofgren, vice president of BrightStor product management at Computer Associates, sensibly notes that: "Traditionally, SRM looked basically at disk, and didn't consider backup at all."