Avoid out-of-storage situations with accurate forecasting: the need for a storage capacity planning capability model.

A study by TheInfoPro revealed storage currently represents the biggest chunk of IT budgets at Fortune 1000 companies. With this much of a financial commitment dedicated to storage, why is it that companies often face emergency out-of-storage situations? How can companies avoid these emergency situations?

Storage infrastructure spending has become a top of mind issue as applications' insatiable thirst for computer storage has grown exponentially. Despite shrinking unit costs, yearly storage costs are rising faster than overall IT budget growth and have become an increasingly large part of the total IT hardware budget, often in excess of 50%. Since storage demand has been too difficult to determine or project due to the size, rate of change, and complexity of storage infrastructures, administrators significantly overprovision storage to avoid shortages. This has resulted in average storage usage of around 40% on non-mainframe systems. This low utilization is what IT organizations need to target and improve, significantly reducing Capital Expenditure (CapEx) spending. Therefore the process that IT organizations employ to determine their storage purchase requisitions is becoming a vital part of the IT planning process.

A Storage Capacity Planning process is a consistent, well-defined IT process of forecasting the amount of storage that will be needed over some future period of time for each application. The goal is to plan new capacity deployments at the right time to meet anticipated demand. The result is the significant reduction and improved predictability of CapEx spending.

A Storage Capacity Planning Capacity Model (SCPCM) is a very helpful tool for organizations to develop a storage capacity planning process that meets their business needs. A SCPCM provides the industry a consistent model to define and communicate:

  Best industry practices for lowering "utilization threshold" for
  earlier out-of-storage alarms
  Where an IT organization's processes are now
  How an IT organization can modify its existing processes
  A roadmap of how to get to a desired state from an existing state
  A way to measure progress

OVERVIEW OF THE STORAGE CAPACITY PLANNING CAPABILITY MODEL

The SCPCM provides a framework for organizing capacity planning processes into four stages that lay successive foundations for continuous process improvement. These four stages define a simple scale for evaluating a storage capacity planning process capability. These stages also help an organization prioritize its improvement efforts. The four stages are: Estimation Planning, Resource Side Planning, Integrated Application Side / Resource Side Planning, and Business Plan / Resource Side Planning.

ESTIMATION PLANNING

In the Estimation Planning stage, IT concentrates its efforts on managing its storage subsystem hardware. The IT organization determines what type of storage to provide; provisions and allocates LUNs to application servers; and determines and implements data protection (i.e. RAID, snapshots, etc.), backup/restore, and disaster recovery methodologies. Based on an occasional inquiry using an SRM or homegrown tool, IT has an estimate of the amount of storage currently deployed and what is currently allocated.

In this stage, the IT organization employs no formal storage capacity planning process.

Based on requests for more storage from application owners, IT administrators allocate LUNs from existing inventory. Somewhat more proactive IT organizations purchase additional inventory in advance of its educated guess on when existing inventory will run out. How far in advance is based on its "storage lead time", which is the amount of time it takes to get storage purchase requisitions developed and approved; ordered and delivered from the suppliers; and then installed, configured, and provisioned. To determine how much storage to purchase, IT relies on one of the following mechanisms:

  The business unit or application owners tell them how much additional
  storage they will need.
  They trust their "gut instinct".
  They make purchase decisions based on what was bought last year, last
  quarter, or last month. Since current storage consumption and
  consumption growth rates are not accurately known, this stage
  typically exhibits very high over-provisioning as storage
  administrators look to avoid out-of-storage situations, thereby
  wasting valuable CapEx budget.

Despite overall over-provisioning, incidences of out-of-storage situations frequently occur when individual application server volumes or storage subsystems run out of storage without the "storage lead time" advanced notice. This over-provisioning typically results in a 30%-40% utilization rate (the amount of used storage divided by the amount of allocated storage). Out-of-storage emergencies are very expensive to the IT budget as well as to the business overall.

RESOURCE SIDE PLANNING

In the Resource Side Planning stage, IT semi-automates the collection of storage subsystem data and forecasts future purchases based on current subsystem inventory trends. This stage does not typically take into account trends in the demand for storage, but carefully manages resource side inventory levels to ensure that overall inventory never runs out.

At this stage, IT employs semi-automated collection of storage subsystem information. The information collected includes:

- The storage subsystem's maximum capacity.

- The amount of raw physical storage that is currently installed on each subsystem.

- Of the raw physical storage deployed:

- The amount usable by applications.

- The amount used for data protection (RAID, snapshots) and accounting.

- The amount used as a shared pool for thin provisioning.

- Of the storage that is usable to applications, how much is actually currently allocated to applications.

- The growth rate of physical storage.

As in the Estimation Planning stage, at this stage IT allocates LUNs to application servers based on requests from application owners. By trending out the inventory of storage that they allocate to applications, IT will develop an estimated date when usable storage will run out. From this data the "storage lead time" will be subtracted, providing a timeframe when additional storage will be purchased. Often storage is purchased quarterly and the amount purchased is based on the historical growth rates of allocated storage.

The advantage of the Resource Side Planning stage over the Estimation Planning stage is that the administrator manually monitors the storage inventory so they are less likely to encounter out-of-storage situations where no additional storage exists and emergency purchases must be made. However, since resource monitoring is not fully automated and application consumption is not monitored, out-of-storage situations on individual application server volumes still occur. More importantly, IT has no idea how much of the allocated LUNs are actually being used by applications and at what rate they are being consumed. Therefore IT has no way of assessing if the incoming storage requests are too early, too late, or right on time. As a result, significant over-provisioning still occurs to the tune of a 35%-45% utilization rate.

INTEGRATED APPLICATIONSIDE / RESOURCE SIDE PLANNING

At this stage, IT systematically collects, stores, and analyzes information about applications' storage consumption and growth in addition to the storage subsystem inventory described in the Resource Planning stage. IT can then produce storage forecasts based on application consumption and then correlate that with the storage subsystem inventory.

The Integrated Application Side / Resource Side Planning stage first analyzes the application side consumption to understand application host volume projections to determine when, where, and what type of storage is needed. The process then correlates application consumption analysis with the storage resource inventory, which allows storage administrators to develop a storage capacity plan based on how the current subsystems' inventories need to be allocated to efficiently meet the application consumption demand.

The strength of the Integrated Application Side / Resource Side Planning stage is that it can successfully answer the five fundamental questions that must be answered in order to run an efficient storage capacity planning process:

1. When additional storage is needed -- deploying too soon causes unused storage to depreciate on the shelf, and it does not take advantage of the 25% per year storage hardware price declines.

2. How much storage is required -- how much storage is required to stay within the target utilization levels.

3. Where storage is needed -- specifically which volumes are projected to approach utilization thresholds and require additional allocated storage/

4. What type of storage is appropriate -- whether expensive high end storage is required, or can lower priced tier 2 or 3 storage fulfill the applications' requirements.

5. Where will the additional storage come from -- the required storage can come from existing unallocated storage, from the purchase of additional disk drives in existing subsystems, or from the purchase of new subsystems.

The Integrated Application Side / Resource Side Planning process is outlined in the following steps:

1. Application side consumption data should be collected at least daily from all volumes on all servers in the IT environment. This data should include each volume's total storage allocation, total storage used, and total free storage. The data should be stored in a historical database for analysis.

2. The IT organization must define the following storage capacity policies:

- "Utilization threshold": balances the avoidance of out-of-storage situations, hardware cost, and performance considerations. Typical target thresholds are 70%

- 80% for most applications.

- "Storage lead time": the time it takes to purchase, deliver, install, configure, and allocate new physical storage.

- "Storage increment": the amount of storage that should be added to a volume when it approaches its utilization threshold. This amount should avoid allocating/purchasing storage too often and allocating/purchasing storage too early. These policies can be unique for different groups or classes of volumes, or standard policies that apply to all volumes in the entire IT storage infrastructure.

3. The application side consumption data gathered in the first step should now be used to determine the forecasted growth of each and every volume using advanced statistical analysis. Once growth for every volume is forecasted, it must now be compared to the utilization threshold to determine when in the future the threshold will be violated. This will result in a list of volumes and dates for when its growth will violate the utilization threshold. From this data, the storage lead time should be subtracted to determine when storage needs to be ordered to ensure it gets allocated to the volume in alignment with when its utilization hits the threshold. From the storage increment, the amount of storage to provide the volume can be calculated.

From the analysis in this step, the following can now be answered:

- When additional storage is needed

- How much storage is required

- Where the storage is needed

4. Each volume that requires additional storage supports an application. Each application within the company has different availability and performance requirements which dictate the type of storage required for each application. By correlating the applications' requirements to the volumes that support them, the type of storage (i.e. expensive tier 1 or less expensive tier 2 or tier 3) that each volume requires can then be determined. From the analysis in this step, the following question can now be answered:

- What type of storage is appropriate

5. With the information determined in the first four steps, the application demand for additional storage is well understood. Now, the application demand must be correlated with the storage supply to understand where the additional storage will come from. The following analysis should now be used:

- Compare the additional storage requirements to the inventory of unallocated storage in existing subsystems. At the appropriate time, allocate the unused inventory to the appropriate volumes. Since there are no purchase and delivery steps, the storage lead time for this case can be shortened.

- If unallocated storage doesn't cover growing application demand, then the next step is to look at the storage subsystem maximum capacities and unused slots to determine if additional disk drives can be added to existing frames to provide the required storage. This may be the most cost effective purchase option.

- If the above two steps still do not provide the necessary storage, then new storage subsystems will have to be ordered. Often a portion of the application data must be migrated from the existing subsystem to the new one. From the analysis in this step, the final question can now be answered:

- Where will the additional storage come from.

6. With an ongoing capacity planning process in place, a storage infrastructure budget can be accurately determined and refined as conditions change. Work force plans to purchase, install, configure, and allocate storage can now be defined and adjusted to form a living "To Do" list for the coming quarters.

At this stage the IT organization can start looking at this complete process as the storage supply chain, a systematic inventory management process for its most consumable asset--storage. This supply chain model incorporates application demand consumption analysis; physical storage supply and unused and unallocated inventory costs; the cost of data protection, backup and restore, and disaster recovery methodologies; storage tiering strategies; cost of capital; human resource management; and timetables for purchasing and deployment. As a result, CapEx and administrative resources can not only be optimized, but also become much more predictable.

BUSINESS PLAN / RESOURCE SIDE PLANNING

The Business Plan / Resource Side Planning stage is exactly like the Integrated Demand Side / Resource Side stage with one exception. Instead of performing application demand forecasting from historical storage application trends, storage demand forecasting is based on Key Business Metrics (KBMs) from the company's business plan.

Using historical data, key business metrics are correlated against storage usage. Once correlation algorithms are determined, storage forecasts can be developed by using KBM values from next year's business plan.

For example, a retail chain correlates the number of new stores it opened last year and last year's revenue to storage usage. From that data, storage forecasts can be generated based on next year's planned new store openings and revenue projections.

WHAT'S NEXT FOR YOUR ORGANIZATION?

The exploding demand for storage and runaway storage CapEx have made the need for a disciplined storage capacity planning process a high priority. When half the hardware budget is spent on storage hardware, it's simply no longer acceptable to have 60% of storage assets unused. As a result, here is a recommended outline for IT organizations to follow to improve storage infrastructure decisions.

1. Determine which stage of storage capacity planning best matches the current processes.

2. Determine which stage of storage capacity planning best fits the company/organizational needs.

3. Define storage capacity planning process and requirements. Identify areas to automate and seek software and service partners who fulfill the requirements.

4. Develop a roadmap with specific milestones to get to the desired state from the current state.

5. Evaluate products from vendors who meet the requirements of the process.

6. Implement and refine through a continuous feedback loop.

Frank Kettenstock is Vice President of Product Management at MonoSphere.

www.monosphere.com