The backup process has a myriad of costs associated with them. There is the obvious cost of the hardware – backup servers, backup storage and network infrastructure. There is also the cost of operations to manage the backup process. One of the biggest costs and one that is seldom addressed is the cost associated with the backup software itself; upfront purchase price, upgrade and expansion costs as well as on going support and maintenance costs.
Understanding backup licensing costs has always been a difficult task and initiatives like server virtualization make it even more so. In this report, Storage Switzerland will explain the motivation behind backup software licenses, review the various models that are in the market today and suggest a new model for the future.
The Purpose of Backup Licensing Models
Simply put, the objective of a backup license model for most software vendors is to extract the maximum amount of dollars from your IT budget without upsetting the IT Manager so much that they begin to look at other options. This creates a relationship that is adversarial in nature; the more important a set of data is to the customer, the more they are charged to protect it. Certainly it is in the best interest of all involved for the backup software company to be profitable, however, licensing models need to change to be fair, especially given the new realities of the data center environment.
The Evolution of Backup Licensing Models
Client Based Licensing
Almost all backup licensing schemes started with a client based model. These licenses charge by the number of and types of clients protected.
For example, if 12 Windows servers and 10 Linux servers needed to be backed up, a 25 pack of clients needed to be purchased, plus a Windows protection license and a Linux license. Any of these servers that had applications running on them would also need specific modules for each application. So if three of the Windows servers had Exchange, Oracle and MS-SQL installed, a module for each of these applications would have to be purchased. Typically applications are licensed per server that the application is running on. To make matters worse, Oracle on Linux often had a different price than Oracle on Windows and the two licenses could not be interchanged.
Licensing was not just limited to clients either. There was also a charge for the devices that you were backing up to and the servers that were transferring the data. Devices were charged by the number of tape library slots or number of tape drives installed. In fact, some vendors charged by the type, speed of the drive and the capacity of the device itself.
Even prior to server virtualization, these license strategies were very complex; with virtualization they became almost impossible to manage. Vendors, for example, started licensing their VMware modules on a per core basis. This meant that license administrators had to count the number of cores in addition to the number of servers and applications.
The obvious problem with this license strategy is that it was incredibly complex to understand, much less to order from. It was also very time consuming to figure out license renewals and software support. These schemes also hurt customer relationships leaving the customer feeling “nickled and dimed”. Clearly a change was needed.
Capacity Based Licensing
Capacity based licensing is based on the amount of data being protected and it is typically the size of the full backup but it can also be the total amount of data under protection (the size of all data copies). The complexity of the above client based licensing model plus the need to keep up with the rapidly changing virtual environment drove this change.
Under this model, the customer did not have to worry about what operating systems, applications or number of systems were going to be protected. They would typically pay a quarterly or annual fee one time based on the above capacity calculation. This was a simpler model than client based licensing but it still had its challenges. First, ascertaining the exact amount of capacity under protection, or more importantly, the amount of capacity that actually needs to be protected, is fairly difficult. Secondly, a capacity based license tends to force customers to pick and choose which data should and should not be protected. The safer choice is to protect everything but it motivated customers to protect as little as possible.
The major problem with the capacity based pricing model is caused by the never ending growth of data in data centers. The good news is the cost to physically store this data continually decreases. The bad news is that data protection companies don’t decrease their prices at the same pace. So while the cost to store data is slightly reduced because capacity costs continue to decline, the cost to protect data increases at a much sharper rate.
Cloud Based Licensing
Most backup providers that leverage the cloud for backup storage use a similar capacity based licensing model, typically providing the software at no charge and only charging by the amount of data stored in the cloud. Depending on the vendor, this can be the gross data set or the net data set after deduplication and compression.
Again, similar to legacy software applications, these vendors charge you for the amount of data that you store in their cloud. As with the legacy model, as more data accumulates in the data center, there is a continuous cost increase of cloud storage.
The Need For A New Model
Capacity based licensing models served their purpose; they simplified the process of acquiring software to protect the enterprise. However, due to the growth of data plus the rate at which that growth is occurring, these models increase costs instead of lowering costs. As with the rest of the IT industry, backup licensing needs to evolve to a utility based model where the service is paid for as it is used.
An example of this can be found in cloud computing where users pay by the processing cycles required by an application rather than for the physical hardware and the underlying infrastructure. Furthermore, they are only charged for the processing cycles consumed by application. Data protection needs to evolve to the same model.
There are two basic ways to design a pay-per-use backup licensing model. You could charge by the backup or by the recovery. Paying for the backup would not make sense. There are too many “events” since backups are occurring all the time. Also we do not want to discourage users from backing up. That’s a good thing.
The real business value of the backup process is evident when a recovery is needed and works. These events are far less frequent than the backups that set them up so accounting for them will be easier to assess. As a result, a recovery based model is the ideal pay-as-you-go use case. The customer is paying for the software when they actually recover data which is also the moment that the backup software proves its worth.
A Recovery Based Licensing Model
Architecturally speaking, a recovery model would start off with an industry average. Typically organizations restore less than 5% of their data in a single year. So a realistic starting point would be to charge at 5% of total capacity. The software licenses for all operating systems, databases and clients would be provided at no charge to the customer. Then on an annual or quarterly basis the license would be adjusted to reflect the actual restores that occurred. This means that backup software would have to add intelligence to track and report on restore activity but there would not be a charge based on number of clients or capacity.
Advantages of A Recovery Model
A recovery model essentially encourages good behavior. The first big advantage to a recovery based model is that it encourages backups. There is no reason not to back up everything in the environment. It should allow for all data to be protected, all the time.
In a recovery model, much of the cost of the backup process is now deferred to the recovery event. This means the cost is tied directly to the business reason that backup is implemented; the recovery of data. Data protection is now tied to a line item on the budget that resulted in something that was important to the enterprise – a return to operations of an application or data set that was down. The investment in backup architecture is now concrete.
The model also leads to more predictable budgeting. While the data set increases, the amount of data recovered remains nearly flat. Recoveries tend to be of the active data set, the size of that data set is not growing nearly as fast as the size of the retained data (old data and copies of data).
Recovery Based Model Concerns
As recovery models begin to appear on the market, there are a few concerns. The first is a very legitimate concern over a disaster recovery and DR testing, where most if not all of the data needs to be recovered. Without special consideration either of these events could be very costly to an organization. Both cases can be handled by an ‘out clause’ where a DR is allowed once per year and DR drills can be pre-declared.
The second, somewhat surprising concern is a “data hostage” concern. Data is being held until it is needed and the percentage of data needed during the year is what determines the cost in the following year. But the price never changes during any given year. Of course this is not the case with the recovery license model as there is an upper limit included to provide organizations with real cost predictability. There is a maximum amount on the cost of recovery that an organization will pay even in the unlikely event that they need to restore all of their data. This model is unlike other storage service solutions in the market today for archiving that charge a very low price to archive data and a much higher price to access the data. In Storage Switzerland’s analysis, the cost per recovery is far less than the cost of capacity based or client based licensing strategies.
The technology of data protection is pretty simple, copy data from point A to point B. The technology continues to advance to allow these transfers to happen more intelligently and more reliably. Also, improvements are constantly being made to improve recoveries including cloud based recoveries. These remain the features that software vendors use to differentiate themselves. But all this investment in technology has led to little change in the way backup software is purchased.
Recovery based models are that dramatic change. They create a fair pay-for-use relationship that should allow IT to better correlate the data protection investment to the business value that it provides.
Asigra is a client of Storage Switzerland