Creating a model for data backup capacity

Creating a model for backup capacity will assist you in forecasting your storage needs for a given budget cycle. This way, you can consolidate your storage purchases and save big bucks.

One of the major benefits commonly touted for implementing a storage network is improved utilization. With proper...

planning and cooperation, you can forecast storage needs for a given budget cycle and consolidate storage purchases, resulting in considerable savings. However, most managers neglect the backup environment, particularly the tape infrastructure. This leads to a flurry of panicked activity, often resulting in a hurried purchase to address the immediate need. Repeating this numerous times results in a chaotic backup infrastructure.

Forecasting your tape drive needs

Once you model your current tape environment and measure your actual data growth needs, you can forecast when your current environment will run out of steam. In this real example, the environment was basically already maxed out. By consolidating backup servers on a common platform, the tape infrastructure could be better used, resulting in extended life before it hit the capacity ceiling.

To develop a model for capacity planning, there are two dimensions of backup capacity that need to be examined and understood--one related to performing daily backups and the other pertaining to retention and recall of data. The former is driven by the requirement that sufficient resources are available to meet daily demand. Think of that as a bandwidth consideration. The latter--relating to retention and recall--focuses on the availability of backup data for timely recovery, and is a policy and media capacity consideration. Let's look at the components of each of these in a traditional LAN-based tape-centric backup environment.

Bandwidth Generally, bandwidth capacity for backups pertains to how much data must be moved within the backup period, and is based on a number of components along the data path. If we look at a traditional network-based backup, the bandwidth capacity is the slowest of the following:

  • The primary disk from which data is read
  • The backup client placing data on the network
  • The network transporting backup data to the server
  • The backup server reading, processing and writing data to the tape storage system
  • The aggregate tape write performance of the tape storage system
While all of these are factors in backup performance analysis, from a backup capacity planning point of view, the last three items are the most relevant. The first step is to understand current resource capacity. Here are the basic calculations:
  • Network bandwidth = number of backup channels x network throughput
  • Server bandwidth = number of media servers x I/O throughput capacity
  • Tape drive bandwidth = number of tape drives x tape write rate per drive
  • Backup window (in hours)
Your current daily backup capacity is the lowest of the first three items multiplied by the backup window. To make this number more realistic, I'd suggest factoring a reserve capacity, based on actual observed performance. A reasonable place holder is 10%.

Next, determine the current utilization rate. This entails measuring the current daily backup volume, both average and peak, and determining the percent of capacity in use. In many environments, a much higher volume of data is backed up on weekends or other periods of high activity. Be sure to factor this into your calculations.

Once this baseline is established, the next step is to estimate the rate of growth. This should include both the data requirements of current applications as well as those being planned in the future. Also, be sure to include other factors such as consolidation of multiple backup environments. Obviously, the accuracy of this forecast is critical to the validity of the capacity planning exercise. So I strongly recommend that these numbers be measured against actual data periodically and the forecast adjusted accordingly (see "Forecasting your tape drive needs").

Retention and recall
The raw capacity requirements for backup are considerably different than for primary data. To begin with, there's much more backup data in an environment. Depending on retention policies, there can be anywhere from five to 50 times more backup data than primary data. A critical question when developing the capacity equation is how quickly this data needs to be recovered.

Backup policies make a difference

This graph illustrates the effect of optimizing backup operations and policies. The company would outgrow the tape library capacity by February 2004 without any change. Implementing changes in processes extends that out to December 2004, and the addition of a second library further extends the viability of the system until past August 2005.

Because tape is a removable media, its capacity is nearly limitless--in theory. Yet from a practical standpoint, limits have to be set related to data retention and the restore time. The key factors under consideration are the size of the tape library, the media required and the costs associated with transporting and storing tape media.

To recover data in a timely manner, a tape library should, at a minimum, have the capacity to retain the current version of all data. That means it should be able to hold the most recent full backups and related incremental volumes. In many environments, the requirement is considerably greater, often targeting an onsite goal of 30 days or more. To calculate this, one must consider how many backup cycles (full backups plus associated incrementals) are required to meet retentions and recovery requirements along with the amount of data associated with each. It's also wise to factor in a realistic estimate of average tape utilization--something less than 100%.

Given this current capacity, you should then determine the current utilization rate and apply the same data growth projections used in calculating bandwidth to determine projected volume retention capacity requirements (see "Backup policies make a difference").

Up to this point, we have considered capacity planning based upon business as usual, making no modifications to the existing environment. Assuming an accurate growth forecast, consider this the high-dollar budget number and use it as a baseline for the capacity planning process. Our experience has been that in almost every environment there is room for improvement, including operational changes that increase tape utilization and policy changes that dramatically reduce the quantity of backup data that must be maintained in a library.

This is the stage in the capacity modeling process to begin to play "what if." Look for inefficiencies in the environment and explore options to improve them. Then rework the capacity numbers and translate them into budget dollars. "Forecasting your tape drive needs" shows an example of forecasting tape library capacity, including some "what if" options.

What if your backup architecture is not the traditional network architecture, or you are considering moving to a LAN-free or disk-based design? The need for capacity planning becomes even more critical because the costs are typically greater. A LAN-free environment usually requires more expensive software licenses as well as higher costs for storage area network (SAN) infrastructure. It can also potentially increase the resource contention for tape devices if it is not planned correctly. A disk-based design eliminates the tape drive resource contention, but still retains network bandwidth considerations. In a pinch, you can buy those additional terabytes of tape media and find a place to store it far more easily than purchasing and integrating an equivalent amount of storage in the data center.

Developing a capacity planning model for backup is an important step in improving backup efficiency. By maintaining and publishing capacity trending numbers, an IT manager can begin to socialize the impact of enterprise data growth on backup.

Dig Deeper on Data storage strategy