Published: 04 Aug 2003
I recently had dinner with four friends who happen to be directors of IT operations. We talked about many aspects of IT, but when we meandered into a discussion on storage, I was surprised to learn that my friends' companies still treat storage as an appendage of a system environment. Each firm has four computing environments: mainframe (IBM zSeries), AS/400 (now the iSeries), Unix and Windows. Each operates as a storage island with its own equipment, policies, procedures and staff. And consequently, several things are true: In this scenario, the way data is protected depends on the computing platform, equipment in one area is overburdened while another lays dormant, and the four groups have unique ways of doing the same tasks. None of that is good.
If you read the typical trade press and vendor rhetoric, you'd think these guys are a bunch of dinosaurs. Aren't companies now sharing storage area networks (SANs) across heterogeneous environments? Haven't many firms built common storage organizations? The answer to these questions is definitely yes, but my friends aren't that uncommon. I estimate that for every tightly integrated storage strategy, there are at least an equal number of firms that still maintain separate storage silos.
My IT director-friends say that their companies are all open-minded about moving to a more centralized, consolidated storage model, but they wondered how to start. I decided to find out how they should proceed.
First, here's what you should absolutely not do: Choose a product as the impetus for a unification movement. Yes, EMC--or any one of its competitors--will sell you SANs and stories all day long about how it can amalgamate everything into one big, happy storage world. To some extent any one of them can, but wrapping a storage consolidation strategy around equipment is like planning your beach vacation around the location with the warmest water temperature. An important detail perhaps, but many other considerations should be higher priorities.
Rather than start with a technical solution, embrace a comprehensive strategy that includes people, processes and technology. Approach storage consolidation as a project with a full life cycle. It should include these five phases: assessment, design, testing, implementation and operations.
Assessment. Before proceeding with technology, it's important to step back and assess the business requirements. Companies need to examine the link between business initiatives and storage, and they should consider where this relationship is headed. When you add up all those operating environments, how much data does the company have? How fast is data growing? Is the company doing anything with reference data? Does the company classify data in terms of its business criticality and confidentiality?
The answers to these questions should provide a general direction. The next step is to do an IT assessment to see how the various storage islands are able to meet business objectives. Chances are that the mainframe and AS/400 environment adhere to more stringent processes than Unix and Windows. Those methodologies can be a model for the other platforms. IT staff for the more mature platforms can also provide training and education.
The final step in the assessment phase is to look at the technology. Little by little, you'll want to reduce your hardware and software vendors down to a few standards. Look across the platforms to identify vendors that offer great products and services with cross-platform options, and plan on phasing out all others.
Design. Knowing your business objectives and having a view of technology and IT processes, you can start designing solutions to meet your long-term needs while reducing operating costs. The overall goal should be a unified storage infrastructure with common operations, but Rome wasn't built in a day. Moving too quickly could alienate staff and put critical data at risk, so it's important to design a plan with realistic goals. A good rule of thumb is to set quarterly objectives that can be measured accurately, and clearly support the overall direction. For example, a first-quarter goal might be to establish standard policies and procedures for data classification across the enterprise. In the future, target larger cost-saving opportunities, such as merging the Unix and Windows groups or sharing 3590 tape libraries across zSeries and iSeries platforms.
Testing. In moving from a platform-centric to a consolidated model, the test phase will be difficult, but it's extremely important. The complication comes from worlds colliding; IT staff will need to learn new technologies and terms, and they will be judged by different metrics. To support different platforms, technologies will need special software drivers, specific revision levels and tricky configurations. Get used to these challenges, because they never go away.
System and storage configurations can be very temperamental, and vendor interoperability matrices are notoriously inaccurate and quickly become out of date. A test bed can uncover and correct these types of problems in the sandbox, rather than in production. You'll want to appoint an interoperability test manager who's comfortable with multiple platform environments, storage technologies and vendor management.
Equipment testing is critical, but don't forget to test processes as well. For example, test the integrity of the backups and off-site media management when the Unix and Windows server teams join forces. Run live disaster recovery tests and judge whether these new groups can meet recovery time objectives and recovery point objectives for critical systems. If they can't, assess whether the issues stem from new processes or poor execution and make the necessary adjustments.
Implementation. Even if this phase is quarters or even years off, it's worth planning for the inevitable. Because the implementation team must be composed of people with cross-platform skills, you'll need to plan for cross training key members of the engineering team.
Get a good understanding of what types of services vendors offer. Do they simply deliver equipment, or do they work with customers to ensure the proper implementation testing and configuration? Be careful here. Many vendors will provide basic setup and nothing else. In other words, the systems will be powered up, but they won't be capable of doing anything without some advanced configuration, scripting or software setup. Be specific with vendors to understand their capabilities and what they will truly deliver. If you aren't satisfied, enlist the help of a third-party service provider--such as GlassHouse Technologies, in Framingham, MA, or North Billerica, MA-based Daymark Solutions--that focus on services rather than products.
Operations. Here, improvements can be made immediately. The key to improving operations is having clear definitions of policies, procedures and metrics up front. Again, mainframe and AS/400 operations provide a great model. The key is to take the positive aspects from across the enterprise without any appearance of elitism.
Remember to measure your results to gauge the overall efficacy of the effort. Some key technical metrics include the number of trouble tickets, mean time between failure and mean time to repair. Those metrics will help IT address problem areas and assess progress. Check out IT business metrics as well. Find out whether there's been reduced capital spending, a return on IT assets, any staff reduction or a declining number of storage vendors.
Share these results with business managers in a clear, nontechnical way to make sure that everyone is on the same page. Consolidation can't be considered a success if IT is giddy with glee while business people worry about whether their critical data is protected.
Combining storage across mainframe, midrange, Unix and Windows platforms can offer fantastic benefits, such as greater data protection, economies of scale and improved services. Many IT shops appreciate this strategy, but they go about implementation the wrong way. Rather than make comprehensive changes, they simply rally around a technology--like a SAN--which they see as being a consolidation hub. These well-intended efforts often result in technological and cultural failures.
Rather than default to technical solutions, companies should approach storage platform consolidation through a full project life cycle, one that includes assessment, design, testing, implementation and operational phases. Be prudent throughout, allow ample time, budget accordingly, and set realistic goals and metrics. This should be a pragmatic process. Following the pragmatic process I've outlined can break down platform walls and lead to productive enhancements in the future.