(1) Think big, start small, scale fast. (2) Automate the repetitive before the complex. (3) Don't automate what you do; automate what you would do.
By Tad Lebeck
Many analyst/consulting firms have dubbed storage automation as the "next big thing" in storage management. For instance, the Enterprise Storage Group ("storage process automation"), the Yankee Group ("storage automation"), and Strategic Research ("storage operations management") have all weighed in with comprehensive reports on the subject.
Yet in a survey of 192 IT managers and executives, conducted by TheInfoPro (www.TheInfoPro.net), automated storage provisioning, for example, ranked very low on the priority scale at both Fortune 50 companies and small to medium-sized enterprises (see charts). It's clear that IT managers are wary of the hype surrounding "emerging" technologies.
While you should be wary of vendor hype (claims of more than 200% improvement in storage utilization simply lack credibility), you can realize significant benefits from automating storage operations. These accrue as automation is incrementally applied to achieve policy compliance and operational efficiency. Storage automation can enable you to approach storage overhead thresholds common in mainframe environments, which are on average between 15% and 20%.
More importantly, automation allows you to
- Set measurable goals;
- Quantifiably improve the process to achieve these goals; and
- Enforce compliance with the policies that drive appropriate service levels.
Automation has its greatest impact, however, when applied outside the domain of storage resource management (SRM). Here, automation can perform and audit execution of data movement, protection, and recovery operations, providing the framework to certify critical data is where it's supposed to be and that it can be recovered according to defined timetables.
Storage managers must focus on realistic goals; the vision of "lights-out operations" will not become a reality anytime soon. On the other hand, real benefits are achievable if you view automation as more than just the provisioning of storage capacity; and think "outside the box" in your approach to implementation.
One approach is to take a page from the way companies have successfully implemented line-of-business applications: Adopt an iterative, process-oriented approach to your automation initiative.
Think big, start small, scale fast—Global implementations are high-risk implementations. Select a portion of a single, well-understood process, and automate, learn, iterate, and evolve. Set clearly defined, quantifiable, and realistic objectives and show success early. For example, start by automating on-demand reporting of system status. This is a straightforward process that does not involve analysis or policy decisions to be made during execution.
Encapsulating the commands and logic for cross-platform data collection and report generation as reusable services can support common requests for information. These can be published to information requestors for use (and re-use) in a self-service model.
Once you've demonstrated initial success and polished the self-service model, follow that by incrementally rolling out services that support policy-based system configuration or data management. Not only will automation deliver improvements in productivity and quality of operations, but it can also provide the framework to enforce and audit policy and service-level compliance.
An example of a "start-small" approach is the case of the retail bank where the operations staff could never remember to perform a simple step to record the configuration of the volume manager before making modifications. This made it very difficult to recover the configuration following failures. By encapsulating and automating the volume change functions and including an initial step to record the volume manager configuration prior to each change, multiple hours-long recoveries were reduced to a simple (automated) process to rebuild the original volume manager configuration.
These clearly defined, encapsulated tasks could be delegated to authorized operators with the assurance that best practices would be followed. The bank was able to build on this by automating tasks at all levels of its storage management operations, disk arrays, file systems, and applications, eventually integrating them to evolve toward end-to-end automation of more-complicated processes.
Automate the repetitive before the complex—Look outside the traditional domain of the storage administrator. Often, these tasks are associated with data movement or protection, or encompass provisioning steps performed by system administrators or database administrators (DBAs), after LUNs have been assigned to a host. Database cloning is a highly repetitive operation that is an excellent candidate for automation.
Sequential tasks to create or copy files for testing or data analysis on pre-provisioned storage infrastructure can be reduced to single operations, eliminating repetitive work for the DBA. Over time, these operations can be extended to integrate on-demand creation and management of mirrored copies or remote replicas to affect multi-site backup-and-recovery practices. Here, automation addresses the more complex task of supporting collaborative operations between storage, systems, and database administration staff and again, provides the framework to certify execution of all local and remote operations required.
Don't automate what you do; automate what you would do—In many cases, automating a current practice merely repeats the sins of the past with greater efficiency. Evaluate and model process improvements that are made possible through automation. Be prepared to evolve and change your policies accordingly. Think about how you would operate if the only way to manage your storage infrastructure were through automation. This brings us back to auto-provisioning. Organizations typically pre-provision storage on monthly schedules—or weekly at best—begging the question: Why automate something I do once a month? The answer, of course, is, you wouldn't.
This school of thought is the major inhibitor to realizing the benefit of a shared storage model. The practice of pre-allocating and dedicating storage to individual applications is an artifact of the operational complexity involved in real-time provisioning. By keeping things simple and containing complex policy compliance, automation allows you to operate differently. If it only takes minutes to respond to a storage request, why wouldn't you provision on-demand? By allocating storage based on need, as opposed to forecast, you can avoid an unnecessary buildup of over-provisioned storage capacity. In addition, you can incrementally reduce re-provisioning thresholds and service response times as confidence in your operational procedures increases.
Typically, automated operations have been handled through scripting. However, automation technologies are available today that provide a significant improvement over scripts. Besides the benefit of substituting vendor-supported software for "home-grown" code, storage automation solutions can manage the distributed process execution required by today's networked storage environments. In addition, they provide a substantially improved process management model that matches organizational realities. This can include a workflow model spanning multiple organizations. It allows administrators to interact with the process, providing policy decisions, when necessary, while the process is running.
Another advantage over scripts is the ability to define a process, delegating its execution and management to enable a self-service model with fine-grained security access control. Most importantly, automation solutions support rapid and iterative deployment, starting with single task automation and evolving to more-complex processes that implement an organization's specific operational practices.
Tad Lebeck is chief technology officer and vice president of engineering and technology for Invio Software (www.inviosoftware.com) in Los Altos, CA.