Policy-based management: Making the most of your storage resources

By Larry Cormier

It used to be that when storage capacity ran low, all you had to do was add another disk array, jukebox, or tape drive onto your network. That is no longer an option. Strapped with increasingly tight storage budgets, IT administrators today are forced to find new, more-efficient, cost-sensitive ways of managing increasingly large pools of data.

One key way is to use intelligent tools to automate management procedures with policies. Storage policies can help with capacity planning and provisioning, allow for non-disruptive transitions, and enable users to control and administer data based on application and business requirements.

What is a "policy?"

When it comes to defining the word "policy," the waters are muddied. While some are quick to say that if you can push a button, then it's a "policy," others are more specific, defining it as a set of configuration parameters that react to computer states or processes.

Policies are not rules, though rules are a subset of policy: Rules are condition/action (if/then) pairs for automated operation and management. Policies span a wide range of parameters from managing high-level business goals and service level agreements (SLAs) to controlling computer rules, scripts, and agents.

Policies may order actions; affect the state of a system, network, or component; or set configurations. They may be either proactive, controlling operations within set parameters, or reactive, responding to events. In an IT world that struggles with staffing issues and storage provisioning, policies can change storage behavior without changing implementation, resulting in improved automation levels and storage consistency.

Policies aggregate multiple variables, which are mapped to business, financial, and operational requirements and needs. Operational variables, for example, may be grouped and mapped to administrative goals. An administrative goal might be to achieve 70% disk utilization across the company's varied disk resources.

Operational variables might include the availability of discovery and reporting procedures, various disk sizes and speeds, storage device types and controllers, and applications and file types served. Administrators could group all of these variables under a single policy setting for 70% disk-based utilization rates, and the policy would implement individual rules to handle operational events within set parameters.

A number of policy-based storage management solutions are available. Data-centric or enterprise-level products provide IT administrators with lasting value.

Data-centric policy

Policy-based management is useful for capacity planning and resource usage, but where it really shines is in providing proactive data administration. This data-centric policy approach allows an organization to manage itself from an information perspective, knowing that storage resources, operations, and practices only exist for one thing: to protect and restore an organization's data.

Since not all data is created equal, data-centric policies allow organizations to set logical data requirements based on organizational needs and application-specific requirements. A best-practice scenario is to set policies at operational levels.

For example, finance companies, which follow strict recording and retention guidelines, aggressively protect their critical data through frequent replication and backup operations.

Secondary and remote servers must store several different copies of the data from various points in time. Inactive data may be automatically migrated to near-line storage, while less-critical data is backed up once a day to secondary offline media.

Controlling multiple replication, backup, and migration policies in these environments is challenging and time-consuming. Policies can order and control data movement according to business requirements (such as certain types of documents) and application requirements (such as database records as opposed to files).

Policy-based management can prioritize the data copies by reserving minimum levels of processing power, guarantee required media capacity, and ensure the accessibility/ availability of other necessary resources.

Policies can also play a key role in disaster-recovery operations in which applications and data need to be restored in a certain order. In a healthcare environment, for example, a bare-metal restore of the database application might come first, with patient database and drug distribution records following.

Policies can range from sending messages to blocking restores or low-priority applications until all high-priority applications have been restored to automating the entire restoration process.

Enterprise-level policies

The Holy Grail of policy-based storage management is the ability to operate reusable policies--regardless of storage type (e.g., DAS, SAN, or NAS).

Central to this is the ability to automate discovery and to present results in both logical and physical views. Once the management software has completed initial discovery and set an update schedule, administrators may institute standard or customized storage management policies to groups of applications, directories, or data. The policies should be able to act and integrate at all levels from a single file or record to workgroup processes to global storage operations.

Doing so can result in more-efficient management of complex storage architectures, an intelligent balance of user intervention and automation, automated policy-based storage features, improved quality processes and SLAs, and more-sophisticated reporting procedures.

Since policy-based storage management provides a more manageable interface and decision pattern, administrators could set policies to handle basic configurations, change issues, and error management. The application could then push the results up to the framework application.

However, there are challenges to policy-based storage management. Customizing policy can be complicated from both the business and technical sides. Customizing policies often require a complicated script-based process, and processing power can be an issue.

Also, while many policies work via intelligent agents, many storage hosts are already overloaded. To relieve the processing load, some storage management vendors off-load agent activity from the operating system onto dedicated devices (e.g., interconnects), but this can add an unwelcome level of complexity to such devices. Some enterprise storage managers use server-based storage management software modules to control and process its application and device agents across operating systems, applications, and storage devices.

Beyond backup, policies can be used to help automate some of the more-challenging aspects of content management such as preserving reusable sections, tracking revisions among different authors, administering security, and centralizing and managing active and archived rich media content. On all these fronts, storage policies, particularly those set from a data-centric or enterprise perspective, bring IT management closer to realizing the goal of "hands-free" management and data protection.

Larry Cormier is vice president of marketing and business development for CommVault Systems and is also a board member of the Storage Networking Industry Association.

This article was originally published on March 01, 2002