There is no doubt that disk-based backup usage is increasing. The technology can provide meaningful local backup and recovery benefits for critical customer data. However, the fact that customers are adopting disk does not mean that tape is irrelevant or unnecessary. The reality is that there are places for both in the data center. Unfortunately, disk vendors often provide misleading information about tape in an effort to discredit the medium and convince customers to pay a premium for additional disk-based solutions. There can be times when an all disk approach is the right model, but customers need to thoroughly review the pros and cons of the available options. Here are four common myths to consider when comparing disk to tape.Myth 1: Hard disk drives are denser and cheaper than tape.
The measure of storage density is aureal density which refers to the amount of space each bit consumes. As discussed in TechTarget's article, Reliability and roadmap: Two more reasons to choose tape, tape's aureal density has been increasing at approximately the same rate as disk, but tape's bit area (e.g. space available for writing data to) is 200 to 300 times greater than disk. Thus, tape is in a great position to aggressively grow in capacity with LTO-8 expected to deliver 32TB. In contrast, disk-drives are space constrained and will have a much harder time delivering equivalent growth going forward. In short, you should expect to see tape's density increasing faster than disk.
The myth about disk being cheaper than tape is a common one that is frequently reiterated. Simply put, this is not accurate. W. Curtis Preston performed a cost analysis in a recent blog post, Get rid of tape? Inconceivable! He determined when storing 20PB of data, tape costs about $11,000 with 2:1 compression while an entry level disk subsystem costs $12M. These numbers speak for themselves.
Some people focus on the cost of standalone consumer disk drives as a proxy for disk cost, but this is not accurate either. In practice, if you want to use a disk array, you need to use enterprise disk drives. These cost about 2x more than entry level drives. (Read Intel's Enterprise-class versus Desktop-class drives to learn more about hard drives.) Furthermore, there is a significant cost associated with other disk array accessories such as RAID controllers, power supplies and enclosures. (RAID controllers are particularly noteworthy since they are unique to each array manufacturer and represent significant value and thus cost.) Simply put, the price of an entry level consumer disk drive is not an effective estimate of disk array cost and comparing desktop drive prices with enterprise tape is not accurate.
Myth 2: Deduplication brings the economics of disk below that of tape.
Some pundits point to deduplication as the silver bullet of backup to disk economics. However, using Preston's example above, if we assume 20:1 deduplication, the resulting disk cost is $600,000 which is still over 50x more expensive than tape! Remember the price assumes a very basic and entry level disk subsystem. It also assumes the disk vendor does not charge a premium for deduplication technology while in practice, most do.
In short, there is no doubt that deduplication technology decreases the cost per GB of disk subsystems. However, deduplication benefits vary and so the actual capacity gained and cost savings is less clear. The economics of tape do not depend on the uncertainty of deduplication and can be consistently delivered across all data types. In virtually every case, tape provides a compelling $/GB benefit over disk storage. Tape also delivers greater scalability, power and cooling benefits versus disk subsystems.
Myth 3: Disk-based replication is always cheaper and better than tape-based disaster recovery (DR).
People talk about using disk systems and deduplicated replication to improve DR. There is no doubt that this technology provides benefits, but it is not a panacea. If you want to replicate, you will need a remote data center and must pay for the power, cooling and space. Additionally, you will need enough bandwidth to ensure that replication can finish within the required window and must frequently upgrade the throughput as data grows. Finally, the biggest costs usually come from the requirement to purchase a second deduplication appliance which is typically the same configuration as the first and associated software licensing. All of these expenses add up. The total cost can be significant.
If you compare the replication model to tape-based DR, you will find a very different cost structure. You only need one tape library or autoloader. Also, you can leverage an offsite tape vaulting partner like Iron Mountain to handle the physical tape movement. There is no need for multiple libraries, remote data centers or expensive bandwidth. In practice, you simply complete your backup to tape or disk, copy the completed backups to tape and then hand the tapes to Iron Mountain at which point your DR operations are complete. The result is a much simpler yet highly reliable DR strategy.
Myth 4: Backup applications always work better with disk.
I am not sure how to interpret statements like these. Backup applications maturely support both disk AND tape. In fact, tape support has been an integral part of backup for a longer time than disk. Also remember most backup software vendors charge a premium for backing up to deduplicated disk and so licensing can be significantly more expensive than tape. Proprietary disk interfaces like Symantec's OST or HP's Catalyst provide tighter integration with disk systems providing greater benefits. However, these solutions typically come at a premium price versus the tape alternative.
The reality is that backup to disk plays an important r
