You've read numerous articles about moving from tape to disk for data backup and you've attended many conferences and webinars about it. But have you really considered all of the implications of this technological change?
We must first define what disk-to-disk (D2D) backup actually means. The most generally accepted definition refers to the replacement of tape media with a disk target. This method is usually transparent to the backup application and end users.
Disk snapshots (point-in-time images) at the storage array level are also considered another form of disk-to-disk backup. Although not technically a "complete copy" in the traditional backup sense, well-planned snapshots offer protection against data loss that's comparable to tape but sometimes it lacks the same level of granularity that many backup software products allow. Some IT practitioners also consider data mirroring a form of disk-based backups. Although it provides a copy of the data, mirroring alone can't be considered a backup because it lacks the point-in-time capability of a backup solution.
Some of the advantages disk-to-disk backup are as follows:
- Configuring your backup software to use a disk target should be a straightforward process provided that you're using a relatively mainstream product. Disk backup is now fully integrated with most popular data backup products.
- If your disk storage array also has data deduplication capabilities, your decision to migrate to disk possibly becomes even more justified.
- If you're planning on reducing tape media handling by leveraging replication to a remote site and you're also using data deduplication at the storage level and/or WAN acceleration, disk-based backups are definitely a good strategy. Security is also a driving factor for wanting to reduce tape media handling.
- Disk storage, especially when configured in a RAID array, is less susceptible to single media failure. Although tape copies offer media failure protection, disk failure in a RAID array is usually transparent to the backup software and addressed more quickly.
- You can run multiple backup streams when using disk without having to worry about availability of tape devices. Of course, you should not expect infinite performance when doing so, but if mount points are an issue, disk will help.
- Disk will typically yield a better utilization percentage than tape. This is especially true when considering tapes that must be sent to an offsite tape vaulting facility at the end of each day, full or not.
Marginal or questionable gains to consider
Performance is often stated as one of the main drivers for switching to disk-based backups. When comparing throughputs, high-end disk can offer a significant performance advantage over tape. However, high-end disk is seldom used as a backup target for cost reasons. The preferred disk type for backups typically only offers a marginal performance increase over tape mostly because it eliminates tape media mount and data seek delays. That said, you must ask yourself if a three-minute tape mount and two-minute first data block seek time will cause you to fail to meet your established recovery time objective (RTO), maybe the backup method is the problem, not the media, and it might be time to start considering replication and high availability.
The disadvantages of disk-to-disk data backup
The cost of disk remains the most significant obstacle when companies consider migrating from disk to tape. Regardless how you look at it, disk still costs more than tape. Disk vendors use sophisticated TCO and ROI tools to help convince customers that disk will hardly cost more than tape but with so much added benefits that you will forget about the cost difference. Likewise, tape vendors use equally convincing tools to prove the exact opposite. In fact, a recent study by the Clipper Group shows that the five-year TCO using SATA disk instead of tape for long-term storage can cost up to 23 times more.
Part of this higher TCO is not only driven by the typically higher cost of acquisition of disk storage (capital cost) but also by the significantly higher amount of energy required to keep the disk spinning and for the air conditioning required to cool the space where the array is installed (operational cost). Essentially, this shows you that if you want to migrate to disk-based backups purely for financial reasons, this probably isn't a good idea.
Data deduplication is now leveraged heavily to increase the capacity of a disk array and this significantly reduces its cost of ownership. However, not all data will benefit from deduplication; data types such as images, encrypted data and compressed files typically benefit very little and sometimes not at all from deduplication. Anyone considering migrating to disk-based backups because deduplication makes it a better competitor against tape should first thoroughly understand the type of data they will store on these devices to avoid costly disappointments.
Although we should never bring up archives when discussing backups (because they are two very different things), archives should be seriously considered before you toss out your tape subsystem. With tapes capable of holding 800 GB or 1.6 TB with an average of 2:1 compression, space becomes the next consideration. Some tape libraries can store more than 700 TB of data within the space occupied by a server rack. With today's cost per square foot for data centers, this shouldn't be overlooked.
Before migrating to disk-based backups, be sure to fully understand why you should do it and the true cost of migration to avoid costly surprise. disk-to-disk-to-tape (D2D2T) may be a good compromise.
About the author: Pierre Dorion is the Data Center Practice Director and a Senior Consultant with Long View Systems.