The Backup Versus Mirror Debate


Having determined that data protection is paramount to an effective disaster recovery plan, it is interesting to note that the industry has, up until very recently, advised that there were only two options for data protection: tape backup and disk mirroring. Tape backup solutions were traditionally aimed at organizations having a large volume of data to back up (capitalizing on the low media cost and high capacity of tape) and adequate " windows " in their processing schedules to allow tape backup operations to be completed. Rarely was there a discussion of restore speeds with tape, for reasons that we will explore later.

On the other hand, mirroring solutions were targeted at organizations that depended on real-time data and whose business processes manifested a high degree of vulnerability to even short duration outages. Often these organizations, which included most of the financial industry, had no available "window" in their operational schedule for performing traditional tape backup. And, for them, time-to-data ”the amount of time to restore data to an accessible state ”requirements were severely constrained.

Many of the arguments made over the past few years by vendors and various commentators for, or against, backup or mirroring are entirely specious. For example, the notion that tape is a dying or dead technology has resurfaced more often than this author can remember and has never been borne out by fact.

Most recently, tape was pronounced dead by vendors of Fibre Channel SANs, who apparently had overlooked the fact that the most frequently cited motivation for fielding a Fibre Channel fabric in survey after survey of customers was to share a tape library.

Others in the industry have announced the death of tape as a function of decreasing cost of disk media on a per GB basis. The argument that tape is being usurped by less expensive disk arrays is flawed in its premise . As Fred Moore, President of Horison Information Strategies, has done an excellent job of demonstrating, most arguments fail to include the system components required to operate tape and disk media, respectively.

According to Moore's analysis, [2] summarized in Figure 9-3, the cost per GB for tape decreased as the ratio of media to drives increased. Taken from the system perspective, disk continues to be 15 to 20 times more expensive than tape for data protection.

Figure 9-3. Disk and tape pricing guidelines. ( Source: Horison Information Strategies, 100 Arapahoe Avenue, Suite 14, Boulder, CO 80302, www.horison.com .)

graphics/09fig03.gif

But Moore and others have also been quick to observe that systemic cost is not the only advantage of tape. The technology also supports an increasingly critical requirement in open systems storage: data portability.

With most mirroring software provided by vendors for use on their high-end disk arrays, only two arrays from the same vendor can be used to form a "mirror pair." "Multi-hop" mirroring, shown in Figure 9-4, is proffered as the ultimate solution for disaster recovery. Multi-hop configurations use a local mirror pair for symmetrical data replication and also establishes a remote connection with asymmetrical replication for geographic dispersal of the data asset. As with mirror pair configurations, all three arrays in a multi-hop mirror must typically be from the same vendor.

Figure 9-4. Multi-hop mirroring with identical arrays.

graphics/09fig04.gif

The important point is that most mirroring strategies are, at root, one-for-one hardware duplication strategies. Veterans of disaster recovery planning will tell you that one-for-one replacement is the priciest way to go in disaster recovery planning for any IT infrastructure component. Recovery on a consolidated platform makes significantly more financial sense.

By contrast to mirroring, tape enables cross-platform data restore (see Figure 9-5). This means that, in the recovery environment, data can be restored to different hardware platforms or storage topologies than those originally used to host data in the production environment. In 20-plus years of disaster recovery efforts, the record suggests that such flexibility is invaluable.

Figure 9-5. Tape enables cross-platform data restore.

graphics/09fig05.gif

Tape is far from dead, and tape vendor roadmaps for their technologies, shown in Figure 9-6, suggest nothing if not a robust and aggressive effort to keep ahead of the capacity of disk media and to continue to provide value to users both as a data protection mechanism and as an archive media.

Figure 9-6. Combined roadmaps of leading tape products.

graphics/09fig06.gif

On the horizon, developments such as O-MaSS technology from Tandberg Data and Imation, and similar tape technology projects at Sony, are expected to boost both performance and capacity into the ranges required by the most demanding environments. Currently, O-MaSS technology is demonstrating capacities in the laboratory of 600 GB with 64 MB per second transfer rates. Plans currently provide for a 2.4 terabyte capacity media featuring 256 MB per second transfer (Generation 3), but the technology, which uses thin-film semiconductor array write heads and a unique optical read system, has the potential for producing media with 10 terabyte capacity and 1 GB per second transfer rates. [3]

Today, mainstream tape media and drives targeted at the midrange or open systems environment offer capacities of approximately 160 to 300 GB uncompressed and transfer rates of between 16 and 32 MB per second. With top-of-the-line automated libraries, tape can deliver a sustained transfer rate of approximately 2 terabytes per hour under optimal conditions.

The sufficiency of these capabilities to meeting the data protection requirements of a specific application must be decided on a case-by-case basis. There is no one-size-fits-all strategy for data protection.



The Holy Grail of Network Storage Management
The Holy Grail of Network Storage Management
ISBN: 0130284165
EAN: 2147483647
Year: 2003
Pages: 96

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net