New remedy for storage headaches: meta-policies ease the pain.

Network policies are rule-based scripts that identify computer conditions, states, and events, and generate appropriate actions. They are mature tools in the data-networking world, where they easily track and act upon standardized networking devices and infrastructures. Storage policies, however, are a different matter. Because of huge number of proprietary storage assets, policy developers are hard-pressed to code policies that can run in heterogenous storage environments. There are also storage realities to observe, such as runaway storage data, greedy database applications that ruthlessly grab disk space, and operating systems that insist on seeing all visible storage as its own affair. Within this demanding environment, network administrators use policies to help them meet service-level agreements (SLAs), keeping storage assets provisioned and configured so that they can meet required space, performance, and availability metrics.

SNIA defines storage-related policies as "the measurable, enforceable, and realizable specification of methods, action and/or desired states that meet service requirements in an storage-based information infrastructure." This means that policies are machine-based reactions that deploy in response to changes in the storage-networking environment. These changes might be driven by events, conditions, or computing processes. Users can measure and enforce both changes and policy-generated reactions from a user interface.

Storage policies may not be as mature as data-network policies, but they are important additions to the storage administrator's arsenal. Policies can manage storage assets to improve and automate backup-and-restore and archiving procedures, supply bandwidth to demanding applications, and assure that a critical backup has the resources it needs.

Meta-Policies

As useful as storage policies are, two factors are inhibiting their development: proprietary, standard-starved storage environments, and a bewildering variety of contradictory policies. Standards and heterogenous environments will help to standardize policies as well, but the second factor requires a different approach. As IT departments struggle to manage growing storage, many of them turn to policies. Many of these policies are specific to individual storage devices, while others are part of storage management applications. The result is a smattering of disparate policies, only some of which actually work together. So storage management developers are concentrating on developing policies to manage policies: policy engines that can manage, maintain, and integrate dozens or hundreds of policies across a storage spectrum. These managing policies have several names in the storage industry, including policy engines, meta-policies and master policies. Meta-policies are object-based, which allows storage administr ators to re-use policy objects among subordinate policies. A single meta-policy interface allows a storage administrator to create and automate sub-policies running across different storage devices and domains.

Meta-policies should handle the difficulties inherent in managing physical objects. Instead of manually configuring each new storage device, storage administrators can create a meta-policy once by applying a set of strictures to classes of storage assets. From then on the meta-policy does the work when IT connects physical assets to the infrastructure.

Policies are heading towards this stage of development, but they are not there yet. Many vendors are developing towards this ideal, but as usual in the computer world, storage vendors define the same terms differently depending on their product architecture and customer focus. But most storage experts agree that meta-policy development is tied around storage resource management (SRM) because that's the natural (read: easiest) place to start, at the physical level. This is true even with so-called application levels, since once policies determine application and data attributes, they work at the physical and logical layers to help meet service level requirements. This leads to a common definition of meta-policies as a set of specific policies that apply to a particular category of storage assets.

In this approach, when IT introduces new equipment into the storage infrastructure the meta-policy directs class-specific policies to operate on the new equipment, such as identifying the asset's capabilities and provisioning and allocating space. Gary McGuire, senior vice president of CA's BrightStor talked about meta-policies in terms of enterprise environments. He said, "We believe that the policies will be invoked when assets are first introduced into the environment, as well as being utilized to ensure the optimal utilization over the life of the asset." Chris Van Wagoner, director of marketing for Commvault Systems, agreed that meta-policies are developing around SRM, but their future usefulness would ideally be much broader. "Meta-policies is a concept that would be broader, if it would pull together what are traditionally different storage management disciplines, into an at-a-glance operational and management approach."

Many storage developers such as CA are working towards meta-policies as a component of enterprise storage automation projects. McGuire added, "Through this class of application, when a new storage asset is introduced into the environment, it is automatically discovered, policies applied, and provisioned into the environment all without human intervention. We are not here today, however, this is the environment that we are working towards. We feel that policy engines are a key component of being able to deliver on our overall vision."

A primary driver for policies and meta-policies is easing management tasks for IT staff. Many storage managers can barely keep up with the flow of storage data, and storage vendors and analysts insist that automating storage management is a necessity for tomorrow's huge data stores. Sophisticated policies like meta-policies will go a long way towards lifting that burden. Van Wagoner said, "Ultimately tools or meta-policies should decide where to put things based on data attributes, whether than letting humans do it all the time. People should be dealing with the strategic issues, the big business issues, as opposed to the physical issues. Then you can start mapping practices to business objectives, and that's where the value is."

Meta-policies are unnecessary with direct-attached storage (DAS) storage environments, since DAS administrators either manually configure their storage or write specific policies for it. But apart from the risk of administrators leaving, if a business moves its storage to a SAN with its multiple applications and devices, policy-based management become extremely important. Some businesses try to get around this by dedicating SANs to single applications, but that defeats the entire purpose of storage networking consolidation and the SAN simply becomes very expensive DAS. In a true consolidated storage environment, Paula Dallabetta, director of marketing at CreekPath, believes that policies are necessary to effectively run the storage network. "In order to run in that kind of environment, you need to have an inherent rule system--one, to meet quality of service agreements and two, to not stomp all over each other."

CreekPath suggests a three-tier policy approach, including explicit policies, rule-based policies and constraint policies.

Explicit policies: These are user-defined and vary according business needs, goals, and strategies. They often characterize application attributes and requirements, and integrate these with established service levels. These rules might direct an SAP backup to a specific storage subsystem, or age data for hierarchical migration.

Rule-based policies: if/then events where a cause provokes an action. For example, if a RAID device goes down the policy will send an alert to the administrator, or if a disk threshold tops 70%, the policy generates an auto-provisioning action.

Constraint policies: Observe the inherent limitations of storage components. Constraint policies would act on the knowledge that SDRF only operates between EMC Symmetrix subsystems, or that a particular application will not write to a certain tape media.

Another primary driver for meta-policy development is automating the service level process. On the theory of "garbage in, garbage out," policies are only as good as the SLAs. Once a business has gone through the challenging process of tying service level agreements to business strategy, meta-policies are vital for measuring and recording success. The SLA process ideally includes:

* Analyze and audit existing computing procedures: IT departments factor in human and machine resources, as well as financial overhead. Metrics will probably be available for some procedures, but daily and weekly averaging will come in handy.

* Consult: IT meets with executives, managers and sophisticated users over their understanding of business goals. IT then works closely with senior executives to clarify and prioritize the resulting goals, and estimates the financial and opportunity cost to meet the ideal SLAs. For example, if two different applications require upgraded systems to increase performance numbers, management may need to choose between them, or scale the upgrades back to accomplish both.

* Establish storage policies: IT bases policies and meta-policies on integrating business goals and SLA decisions. Meeting service levels may involve such projects as upgrading servers and storage, moving to storage networking, planning for data migration, and so on.

* Monitor service levels: IT uses tools such as meta-policies, SRM applications and metrics to track service levels and SLAs. The reports yield specific data that will help refine network storage policies and help generate enforcement procedures.

As storage management grows more critical and challenging, IT administrators are looking to software tools to help them create and manage multiple policies. Meta-policies add value by governing sub-policies, making it possible to re-use data objects and cascading policy changes through the hierarchy. They also ease storage management burdens and allow policies to operate reliably and consistently across storage infrastructures. By helping storage administrators to harness their storage assets, meta-policies can lead to better asset utilization, more reliable capacity planning, and improved strategic management decisions.

www.ca.com

www.creekpath.com

www.commvault.com