Making disk-based backup work

Adding low-cost disk to a backup environment can ease a host of backup woes. Here's how to find the disk-based backup approach that works best for your shop.

This article can also be found in the Premium Editorial Download: Storage magazine: A look inside Hitachi's TagmaStor high-end arrays:

Does your backup software do disk?
Here's how three leading backup software applications deal with disk.

IBM Tivoli Storage Manager (TSM) was designed to include disk as a key element, using a hierarchical model that maps directly to the "disk as cache" approach. The normal process in TSM is to back up to disk and then migrate to tape in an automated manner. Multiple read and write access is permitted to disk storage pools. High-water marks identify capacity boundaries that will force data migration to the next level in the storage hierarchy (usually tape). Data is migrated from disk to tape on a scheduled basis outside the normal backup window. Because TSM is optimized for a continuous incremental model, hybrid backup approaches aren't applicable. Other disk-based approaches may be used as well, but none have the same degree of built-in support.

Legato NetWorker has supported disk-based backup since Version 5; its file device type supports writing backup savesets to disk in essentially the same format as it writes to tape. Version 7 significantly enhances disk support with its DiskBackup Option (DBO), adding an advanced file type that allows simultaneous access. So, for example, a backup operation and a restore operation can occur at the same time on the same device. NetWorker also fully supports staging to disk with automatic migration to tape.

eritas NetBackup has supported disk as a backup device since version 3.1; Version 5 added a new device type known as a Disk Staging Storage Unit (DSSU) that enables staging data to disk and migrating to tape as an automated process. Version 5 also added the ability to create synthetic full backups--an important capability when disk is used to increase the number of incremental backups between fulls, but full backup tapes are still needed to send off site. NetBackup also allows simultaneous access to disk devices.

For those struggling with nightly backups, the arrival of low-cost disk promises to be the greatest breakthrough for improving the process since the introduction of centralized, networked backup in the early 1990s. Done properly, incorporating disk into a comprehensive data protection plan will improve the performance of backup and restore, reduce day-to-day management problems and provide an enhanced degree of data protection. To realize these benefits, it's necessary to answer these questions:

Where disk can help
In survey after survey, users cite the same reasons for adopting disk as a backup medium: faster nightly backups, shorter restore times, better reliability of backup and restore and better overall manageability. Here's how disk can help realize these efficiencies.

  • Faster nightly backups. The impact of disk on individual backup jobs can vary dramatically. There are many potential causes for bottlenecks in the backup process, many of which are unrelated to whether disk is part of the picture. If these issues are causing your backup problems, just adding disk won't help. However, disk can help most with performance problems, such as queued jobs unable to start because of resource constraints and file servers that have many small files that can't keep a tape drive streaming.
  • Shorter restore times. In many environments, the majority of restore requests are for individual files. Recovery performance for a small number of files is usually not a problem in a tape environment--assuming, of course, that the actual tape media is available. The time to load, seek and recover a file from a tape, while certainly greater than with disk, is often within acceptable limits.

Where disk does help significantly is with recovering a large numbers of objects, such as a full file system restore, or in a disaster recovery (DR) scenario where data is likely to be stored on multiple tape volumes.

  • Improved reliability. One of the biggest complaints about tape is reliability. Tape isn't inherently unreliable, but because of differences in how it's managed, stored and handled, there can be wide variances and little predictability. This uncertainty is the real point of frustration. Disk, through RAID, can at least provide protection against media failures. Disk also benefits--from a reliability standpoint--from not being handled and transported.
  • Better manageability. This is the area where disk truly has the potential for the greatest impact. Many of the everyday problems of traditional backup are related to the idiosyncrasies involved in managing tape and tape devices. The failure of one tape device can have a cascading effect that impacts a number of backup jobs; managing and tracking tape cartridges is a process that invites all sorts of glitches. Disk, if integrated correctly, can smooth the process enormously, resulting in better reliability and faster backups and restores.

It's important to approach disk-based backup with a solid understanding of how disk impacts each of these areas in order to realize the potential benefits.

Checklist: planning for disk backup
Successfully introducing disk into your backup process requires a reassessment of your environment. Here are some of the basics to consider:
  1. Review retention and recovery policies. Keeping everything forever doesn't work with disk. Assess current policies to ensure they meet business requirements, including retention periods, backup frequency, recovery time objective and recovery point objective.
  2. Size your backup environment. Capacity will be a major factor in how you decide to incorporate disk. Get an accurate count of how much data exists on tape today; determine which servers and applications are the major consumers of tape and consider their growth.
  3. Understand the impact on backup software. Find out how your software can leverage disk, and if there's any impact on licensing or functionality.
  4. Gauge the impact on the systems. Decide if backups will run over the LAN or SAN, and if additional hardware (e.g., dedicated HBAs or NICs) is needed.
  5. Consider the impact on the network. Bandwidth and port capacity are key considerations.
  6. Know where you'll still need tape. Tape is likely to continue to play a critical role in the data protection hierarchy.
  7. Determine how system administration will be affected. It's likely that roles will change and new areas of expertise will need to be developed.
  8. Weigh how users will be affected. Schedules will likely change, and scripted procedures may have to be changed to accommodate specific applications.

Integrating disk into backup
There are several approaches to implementing disk-based backup (see "The disk-based backup landscape"). Selecting the right approach for your environment is a matter of determining the right balance of functionality, complexity and cost (see "Checklist: planning for disk backup").

Just as there are many benefits in a tiered data protection and archive storage strategy, getting there requires serious planning in the following areas:

Introduction of additional complexity. In most of today's backup environments, restores are simply a matter of locating the tape with the required data. As long as the backup was successfully completed, the data can be located on tape. Depending on the disk-based approach used, the data may be on disk for minutes, hours, days, weeks or even months. This introduces the need to ensure that data is managed and migrated properly. Some backup products have been enhanced to help perform this function, but in practice this represents a new set of daily operational and management tasks that must be performed and monitored. Imagine discovering at the last moment that a nightly backup to a disk cache never migrated properly to tape, and the next night's backup cycle was about to begin! Using disk also requires more disciplined planning--especially if the disk is shared among multiple media servers.

How your backup software handles disk will greatly affect your disk-based backup implementation. Backup vendors have optimized their applications to utilize tape effectively and many base their licensing fees on criteria such as robotic tape libraries and the number of tape drives (see "Does your backup software do disk?"). Disk-based backup disrupts that model to some degree. Key considerations include:

  • Additional software licenses or add-on components required to support disk
  • Restrictions or limitations imposed by the application on the use of disk
  • If the application manages disk space automatically; the type of increments it writes to disk; capabilities for extending capacity when file limits are reached or when recovering and reallocating space as data expires
  • Additional architectural and procedural changes required to fully realize the benefits of disk

Optimization. It's rare to find a data center that's 100% satisfied with its backup/restore environment. A new disk-based backup environment that emerges from a poorly functioning tape-based backup environment is likely to inherit those same inefficiencies. From a performance standpoint, think of a car that can't exceed 60 mph when the speed limit is 70. Putting that same car on the German Autobahn with no speed limits won't make the car go faster. Consider the processing overhead of doing incremental backups of a system that has millions of small files. The time for the backup software to scan those online files and determine what has changed will remain constant, regardless of whether the backup is going to disk or tape. The first step toward better performance is to find the bottlenecks in your backup environment and ensure that the backup architecture can support driving faster disk or tape.

Scheduling and policy. After adding disk to your backup environment, you may need to make some modifications to your current backup schedule. If performance gains are notable and the backup window becomes less of a concern, you have an opportunity to modify the current policy to potentially do full backups more often, or perform differential backups instead of incrementals.

Conversely, you may determine that better resource utilization and backup performance can be attained by doing fewer full backups and more incrementals. However, this would adversely impact recoverability time from cloned off-site tapes. To avoid this, you may want to perform synthetic fulls, where the backup software automatically (on a predefined schedule) generates a new full backup tape from a series of (disk-based) full and incremental (level zero, level one) backups.

The disk-based backup landscape
When integrating disk-based backup, the critical decision is what approach--or combination of approaches--is appropriate for a given environment. Here's a list of the strengths and weaknesses of the various techniques.

Consistency of backups. Fuzzy backups--backups that are not from a single point-in-time--are sometimes used in sophisticated environments that have numerous application interdependencies. While this topic is usually more relevant to DR, it's still important to ensure the data being backed up is useable and in a coherent state. Disk-based backup doesn't address this issue. If a disk-based backup now takes one hour instead of four, it still isn't from a single point in time. Approaches such as snapshots and split mirrors address this specifically, and still need to be used regardless of the ultimate backup approach. Ironically, if snapshots or split mirrors are deployed in a disk-based backup environment that still uses tape for long-term archival and off-site rotation, the backup environment now has a total of three tiers of data protection storage (online disk, backup disk and tape).

Capacity planning. In a tape-based environment (see "The case for tape"), there are several important considerations when determining the required number of tape drives, amount of tape media and slot capacity of the tape library(s). Research shows that for every gigabyte stored online, there are between 10 and 15 copies of that data stored offline. Capacity planning for backup is an inexact science that's further complicated when online storage growth isn't accurately forecasted.

These issues take on a new twist in a disk-based backup environment. Instead of tape media capacity, you need to consider the file system and LUN sizes of your backup environment. Instead of tape library slots, you need to look at overall array capacity. Rather than estimating the compression ratio you may achieve during your tape backups, you need to investigate the size (including RAID level, amount of cache and disk drive) of the disk-based backup hardware, as well as how long the data being backed up will remain on spinning disk. Adding more tapes to a tape library for additional backup capacity is fairly straightforward; adding storage to a disk-based backup appliance requires considerably more effort and preparation. Also, don't forget that buying disk to handle a last-minute capacity crunch is much more difficult than buying some additional tapes.

Networking issues. Numerous approaches are used with today's popular tape-based backup environments. Some back up over the network, with Gigabit Ethernet quickly becoming the norm. The backup servers are sometimes dedicated, sometimes shared and are rarely sized appropriately specific to CPU, memory and backplane. Other backup environments use LAN-free backups, which have the application server backing up directly to tape devices on a storage network (e.g., Fibre Channel). And a few are even investigating server-free backups, which don't involve the application servers whatsoever.

The case for tape
Adding disk doesn't necessarily mean eliminating tape.

And don't forget those other backups, including the network data management protocol (NDMP) and specialized approaches for network-attached storage (NAS), block, file and e-mail-based approaches, hot database backups which require other data structures such as archive and redo logs and other replication approaches that may use iSCSI or even FTP for file transfers. Why is this important? Because with a disk-based backup environment, it's necessary to consider and integrate these approaches in the same way a tape-based backup environment does. The storage for a disk-based environment can be a storage area network (SAN)-based storage device or a LAN-based NAS device. The process of ensuring appropriate bandwidth to the device and allocating and managing disk storage for backup can significantly affect network planning.

DR. Operational (day-to-day) and DR (full-site failure) approaches are often linked, especially if tape is incorporated into the overall DR strategy. Disk-based backup approaches insert a unique kink into DR strategies, which may include off-site rotation of backup and/or special DR tapes.

Portability of tape media is probably the biggest asset of a tape-based approach when the ultimate DR site is "to be determined" at the time of the disaster. Conversely, assuming the disk-based backup array is not affected during a DR event, shipping a backup array (or disks within the array) is questionable at best. Real-time or near real-time replication of a disk-based backup array to another site is also an option, but that requires a dedicated DR site and WAN capacity, as well as the additional expense of redundant disk to protect the primary site's backup disk.

For most environments, tape will continue to play a major role in DR. Disk-based backup can be a key factor in improving the off-site tape production process. Many organizations today don't follow best practices with regard to daily "tape cloning" for off-site storage simply because they lack the time and the number of tape drives needed to meet the demand. Disk reduces the number of tape drives needed, and because multiplexing isn't necessary in this scenario, data can be recovered from off-site tapes more quickly and easily.

Putting it all together
The future of backup is with disk. Just as today the thought of doing non-centralized backups on individual servers is an anomaly in any IT shop of reasonable size, in a few years, backing up directly to tape will also become a rarity. Managing the transition is today's challenge. Storage managers need to consider what to change and how to do it, and make those changes in the least disruptive manner possible. Having a clear understanding of all the moving parts will help you smooth the transition and realize the true benefits of disk-based backup.

This was first published in September 2004
This Content Component encountered an error

Pro+

Features

Enjoy the benefits of Pro+ membership, learn more and join.

0 comments

Oldest 

Forgot Password?

No problem! Submit your e-mail address below. We'll send you an email containing your password.

Your password has been sent to:

-ADS BY GOOGLE

SearchSolidStateStorage

SearchVirtualStorage

SearchCloudStorage

SearchDisasterRecovery

SearchDataBackup

Close