Tape is Not a Panacea


The tape-centric approach to data protection has several potential drawbacks that prevent it from being a universal solution or panacea. These may include:

  • The (comparatively) slow speed of data restore from tape to disk: Currently, tape-based restore speeds of about 1 to 2 TB per hour are possible using high-end automated tape libraries. This may become increasingly problematic for organizations whose data storage infrastructures are approaching petabytes of stored information. However, even in large shops , tape-based data restore speed is a problem with many fixes, however, including the pre-staging of the large percentage of non-changing data on the recovery platform in order to minimize the amount of data that must actually be recovered from tape.

  • Unpredictablity in tape backup completion: Tape efficiency is a function of several factors that must work together to provide consistent and predictable performance. To enable 30-MB-per-second data streaming rates and to operate tape drives at top speeds, the right backup software and its careful tuning are required. Because of the nature of disk (the source of data to be streamed) and disk interface technology, and more specifically because of the processing stack that must be traversed to drive data from disk to memory then out through an HBA to an external tape library, it is necessary for backup software to initiate, not one, but as many as 16 parallel process threads in order to obtain the desired streaming throughput. Often these parallel threads provide the necessary performance at first, but over time, and as a function of the different containers sizes for disk-based data and their tendency to mitigate the balance and uniformity of threads, thread efficiency falls off and performance suffers. Moreover, as the data feeding each stream is exhausted, the number of threads that are streaming to tape begins to drop off, and backup performance also declines. A reduction in performance may also result from an increase in traffic across a network. As networks get busy, and threads become elongated, streams become slower. Whatever the root cause, storage administrators soon discover that only a portion of the number of backups that need to be made have actually been completed within the allowed timeframe. Solutions to this problem vary, but a promising approach is to create a virtual tape library consisting of a disk array that emulates a tape library subsystem. The benefit of this strategy, which is advanced by companies such as Alacritus Software, Veritas Software, StorageTek, and others, is that it capitalizes on the transfer rate of disk (about 80 MB per second), which is three to four times the transfer rate of tape drives at 16 to 30 MB per second, to enable more work to be done in less time. Using virtual tape (sometimes called tape-emulating disk or disk caching) expedites backups and enables, because of additional intelligence in virtual tape software that tunes performance to adjust for changes in data and network load, more backups to be performed more reliably within the same timeframe.

  • The relative disparity in tape and disk drive media capacity: Until very recently, tape has lagged significantly behind disk in terms of storage capacity improvement, necessitating that more tape media be used to store data copied from a single disk drive. Since the beginning of the new millennium , this gap has been closed and tape technology appears to be keeping pace with disk drive capacity expansion.

  • The growing problem of tracking and managing recorded media: The data explosion confronting many organizations has created an increasingly complex media management problem for organizations utilizing tape backup. Over time, and in certain organizations, labor costs for media management may increase the systemic cost of a tape solution to a point where it competes less effectively against the cost of disk to disk. Software is available or under development today to facilitate improved media management.

  • The complexity of sharing and scheduling tape storage devices: The rise of n - tier client/server and other distributed computing models has seen the dispersal of data that needs to be included in a backup scheme. In turn , this has introduced two problems: (1) finding ways to share tape targets, and (2) finding ways to schedule the use by multiple initiators. These problems are being addressed through improved network and fabric attachment strategies for tape libraries to facilitate sharing, and improved tape virtualization techniques to front end tape with disk caches (to improve performance) and schedule management smarts. An example is StorageTek's SN6000.

  • The logistical vulnerabilities of tape-based data recovery: in most cases, tape-based backup schemes involve manual, human- intensive , media gathering and transport requirements that are prone to error and expose data to potential loss before it reaches the recovery setting. As a portable medium, tape is, by design, exposed to more threats. There is no easy solution to this problem, only expedient ones, such as electronic vaulting (see below).

The above are some of the most commonly cited challenges confronting traditional tape-based data protection schemes. Some of the inherent time-to-data challenges of tape are exacerbated by the requirement to handle tape in and out of the primary facility to an off-site storage facility, and from the storage facility to a remote recovery center or hotsite, where data recovery actually occurs. As shown in Figures 9-7 and 9-8, these activities add to the time-to-data delays associated with tape-based data restoral.

Figure 9-7. A traditional tape-based data protection strategy adds substantial time-to-data.

graphics/09fig07.gif

Figure 9-8. Tape-based restore is a function of media transfer rates plus handling and transportation times.

graphics/09fig08.gif

However, some of this delay can be effectively addressed through the use of electronic vaulting and remote tape mirroring techniques that are becoming increasingly available from off-site storage and hotsite service providers (see Figure 9-9).

Figure 9-9. Electronic vaulting technology augments traditional tape to reduce time-to-data.

graphics/09fig09.gif

Bottom line: Tape backup confronts many challenges, but none that are insurmountable given the interest and budget to optimize the technology and the processes surrounding it. It must also be said that, while there are an increasing number of ways to optimize tape for data protection, the critical "time-to-data" sensitivities of some applications continue to drive many companies to the data mirroring option.



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