Opinion

posted on 08 July 2008 13:19

An opinion written by Jeff Graham, senior product manager at Overland Storage.

Virtual Tape Libraries (VTLs) are more frequently being implemented to solve the two biggest problems that users typically experience with their backup; the backup window, and restore reliability. Jeff Graham, senior product manager at Overland Storage explains that prior to selecting a VTL, users must take a structured approach to determining their requirements

A VTL can solve the backup window problem by providing much better performance than a similarly priced physical tape library because it can emulate many tape drives and multi-stream backup jobs. Of course, disks are more expensive on a £ basis and so most people use a VTL as a temporary holding space for backups which will eventually get moved to tape. So, performance and capacity optimization are important attributes for a VTL.

Restore reliability is improved by utilizing a VTL because it is not susceptible to the failures common of tape and tape drives. Most frequently, restores are done from recent backups, so a VTL can be the source for the majority of restores. However, as the need to restore data from backups that were performed weeks in the past, having a VTL with the capacity or optimization technology to store many weeks of data becomes more important.

The selection of the most appropriate VTL for the users depends on these factors;

1. Amount of data to protect.
2. Performance Requirements.
3. Backup service levels / policy.
4. Retention objective.
5. Flexibility.

Amount of Data to Protect

The amount of data to protect is usually a simple figure to obtain because most users already have a backup process in place and have a known data set. With this figure known, the end user can then look at VTLs which can meet the capacity requirements. Users should consider the advantages of using compression to effectively double the disk space of the VTL. There are two cautions with compression, however. Software only compression can greatly decrease performance, the main reason for considering a VTL. For this reason, a hardware compression approach is preferable. Also, some vendors will charge an additional license fee for VTL expansion arrays that utilize compression and other vendors do not. The user should take this into consideration when selecting a device because it is very likely that their backup data set will grow and expansion will be added sometime in the future. Compression ratios will be approximately the same as they are with tape compression.

Performance

Performance requirements can be easily determined by taking the amount of data to protect and dividing it by the backup window. For example, if you have 2TB of data to backup in a 6 hour window, you need 2TB/6Hr » 100MB/Sec performance from your VTL (and your backup server). This calculation should be done for both full and incremental backups and should also consider data growth for the next several years. With this number in hand the user can look for both an interconnection method and a sustained performance that will meet their requirements. For connectivity in the VTL space, the choices are iSCSI or Fibre Channel – though a device that is capable of upgrading from iSCSI to FC offers the ultimate in flexibility.

Performance figures quoted by vendors are typically derived in best case environments and users should be sure to evaluate the performance capabilities of the other components in their backup infrastructure to ensure that they will be able to meet the backup performance target once they implement a VTL. Further, some of the deduplication technologies rely on an on-disk index (data dictionary) which grows as more data is stored on the device. For these devices, performance will degrade over time and manufacturers will sometimes quote two performance figures, the lower figure being the one that the user should use for comparisons. See more about data deduplication in the section on retention.

Backup Service Level / Policy

Users should evaluate their current backup policy to see if they can modify it in order to take advantage of the benefits of a VTL. For example, some companies have a policy of creating full backups each night because they can’t accept the risk of relying on an incremental or differential backup on tape. When using a VTL, restoring from an incremental or differential backup becomes seamless and therefore the user may be able to change their policy for the better. Also, because restores from a VTL are nearly instantaneous, they may be able to offer their customers a much improved restore service level.

Data Retention

Retention objectives can simply be translated as a multiple of the amount of data to protect. If you need to keep 4 weeks of 2TB backups, then you would need 8TB of usable VTL space. While you can increase your retention with compression, this brings up two other technologies – thin provisioning and data deduplication.

Thin provisioning is the notion of not allocating storage to a device until it is actually needed. In the case of a physical tape, there is no choice of provisioning – the tape has a fixed capacity. With a VTL, however, there is an opportunity to “release” any unused virtual cartridge capacity to other devices and get a much better utilization of the available disk space. There is really no reason to carry over the inefficiencies of fixed cartridge capacities from the physical to virtual tape, so it is advised to look for a VTL that can perform thin provisioning of virtual tape in order to make the best use of the available capacity and maximize retention of data.

Data deduplication is a relatively new technology that is getting much interest in the backup space because it has the potential to greatly reduce the amount of physical disk space needed to store many sets of backup data. Data deduplication compares incoming backup data to the data already stored and, rather than saving duplicates of already stored data, it simply stores a pointer. This technology is extremely beneficial for users looking to keep data on disk longer – meaning to increase their retention period. It does come with a cost – both in terms of dollars and in the performance penalty that the deduplication algorithm exacts. It is best practice to look at both traditional (non-deduplication) and deduplicating VTLs to see if your requirements can be met without the additional expense and performance penalty of data deduplication.

Flexibility

Some of the VTL offerings on the market have a great deal of flexibility to be configured as multiple VTLs as well as stand alone virtual tape drives and disk LUNs. This flexibility is important as it is likely that not all the data in any organization will be protected with one approach. Consider, for example, an organization that wishes to use a near continuous data protection (CDP) solution like Backup Exec’s CPS functionality for some of their data, but use the traditional backup functionality in Backup Exec to protect the remainder of their data. In this case the user will need both disk LUNs for the CPS protection and a VTL for the traditional backup protection. A single VTL with flexibility to expose disk LUNs simultaneously will allow the user to make the most of their single VTL investment.

There are some application specific backup utilities that can not address a library changer, and in those cases it is of course very important that the VTL also is able to be configured as stand alone tape drives.

Summary

Users who take a structured approach to determining their requirements and then selecting a VTL that meets those requirements can look forward to tremendous rewards. Satisfied users of VTLs are able to see as much as a 75% reduction in the number of hours spent managing backups. Users typically see a tremendous reduction in costs associated with the backup process as well, such as a 50% reduction in physical tape media spending. Finally, since the physical tape library and media are only being used essentially for archive, that equipment tends to last much longer than if a VTL were not implemented.

^{[Jeff Graham, senior product manager at Overland Storage]}