| QUESTION & ANSWER |
Four contributors advise on users' SAN management wishes, challenges, SAM |
 |
By SearchStorage
03 May 2002 | SearchStorage |
|
|
Which five storage network administration tasks do today's sys admins wish could be automated? What is storage area management (SAM) really all about? Or, how best should you go about developing general storage management policies for your organization?
Our four contributing experts shared their insights recently with SearchStorage in our Storage Management discussion day event, April 18, 2002. Here, we present some excerpts of their discourse. For the full exchange, scan forum posts #362-394 in our Storage Management Discussion Forum.
Contributing experts:
- Bruce Backa, CTO of NTP Software and a SearchStorage expert on storage management issues.
- Jim Booth, director of systems engineering, CreekPath Systems and SearchStorage expert in several areas, including storage management, storage administration and backup/recovery issues.
- Chris Hyrne, vice president of marketing, Invio Software.
- Steve Toole, vice president of marketing, Precise/WQuinn and a frequent SearchStorage contributing expert in the area of storage management.
What is storage area management (SAM)?
Jim Booth: SAM as defined by Gartner is "the centralized management of resources and data across a storage domain(s), providing shared services to a group of servers and the applications." The components of a SAM are as follows:
- Service Level Management
- Policy-based Management
- Workflow Management
- Storage Provisioning
- User-Configurable Automation of Processes
- Auditable
- Security
- Element Discovery
- Element Monitoring
- Event Management and Analysis
- Visualization
- Reporting
A complete SAM solution will have all these components and allow easy usage from a single interface. It will also be flexible enough to automate as little or much as needed in an environment.
Chris Hyrne: Talking SAM is a bit like boiling the ocean. Gartner's definition has just about everything (except data management) rolled into one. I don't expect we'll see the single-console SAM solution anytime soon. Besides, there are a whole lot of (different) tools doing pieces of SAM already installed in shops today. I don't think these are going away anytime soon. I think the focus of SAM will evolve to: How do I manage the interaction between these tools, their administrators and the infrastructures they manage? From a practical implementation point of view, SAM boils down to a workflow problem. From a technology point of view, an interface problem. As a category, it does hone in on application-focused, end-to-end management -- but it's a bit unwieldy.
Please provide some real-world examples of policy-based management.
Bruce Backa: Wasn't it the white rabbit in "Alice in Wonderland" who said, "When I use a word, it means exactly what I choose it to mean, nothing more and nothing less"?
Policy-based management means that I can create a rule (a policy is a rule), and the system or the technology will automatically enforce that rule. For example, you could create a storage quota: "All users: 500 meg." Such a policy would impose a 500M Byte limit on the private storage of all users. Another such policy might be: "Max bandwidth per connection: 5%" (no single connection can use more than 5% of the available bandwidth).
To some degree, policy-based management is to system management as SQL is to databases. In the old days every database product had its own language and syntax. That made building general purpose applications almost impossible. Systems management technology has similar characteristics. Every product has its own unique mechanism and language for configuration. The move to policy-based management is an attempt to normalize the systems management dialogue and give us an easier way to define what we are trying to accomplish.
Jim Booth: A policy-based management system should do the following:
- Collect data from all devices within the storage network (servers, HBAs, switches, directors, storage controllers, and disk devices)
- Compare the collected data with the rules to determine if there are any boundary conditions
- Take action based on a boundary condition (such as, adding storage to a file system before it is full)
Chris Hyrne: A policy engine should function in an advisory capacity. Whether the engine has the capability to "drive" the infrastructure in an automated fashion or not, human intervention should be required to select or approve the action taken as a result of a boundary condition (or scheduled event).
What are the most common storage management problems?
Jim Booth: The biggest problem I see in customer environments is the lack of standard processes (or, processes that are repeatable and defined in line with the business objectives of the company). It is very annoying when the only person who knows about a storage request (and the provisioning constraints) is on vacation in Costa Rica. It would be nice (and is becoming essential) that the processes be built into auditable workflows that do not depend on a single individual or group. I envision DBAs and other application administrators doing their own storage provisioning in the future. This will allow the storage administrators to concentrate on infrastructure issues.
Bruce Backa: The most common issue that I see in pretty much every (large) environment is disjointed or disconnected systems. Some of the storage is RAID, some of it is SAN, some of it may be NAS. Every one of these is different, and managing them as a coherent whole is very difficult. What happens, then, is that you get islands of excellence (areas that are well-managed) in seas of difficulty (areas that are unmanaged). The biggest challenge is leveling that out and having a consistent level of service and management across the board.
Chris Hyrne: It's not that just one individual is critical to successful process execution -- usually getting something done on behalf of an application takes three or more people who "own" bits of the overall effort required to achieve the end goal. In Jim's provisioning case, [this would be] a storage admin to configure and present LUNs, a sys admin for file management and backup, and a DBA to grow the DB.
Any other general advice or input on storage management?
Chris Hyrne: [We] just completed interviews with 75 enterprises [and] came back with the following top five automation requests -- Other views?
- Creating, cloning, expanding or migrating a database
- Including new volumes in an applications replication and backup service
- Capacity management and reclamation of space
- Provisioning of storage, fabric, files and volumes integrated with apllication (DBA self-service)
- Migration of DAS to SAN
Steve Toole: [Editor's note: Steve offered a few posts with rules to follow for setting storage management policies. The first of these posts is excerpted in this answer. Please refer to his other forum posts for more advice.] Network storage policies need to be let gently out of the bag. Have the IT department or IT storage committee communicate with department heads and their employees about the storage policy and the way it will be carried out, especially what housekeeping tasks employees are asked to do. The IT department needs to present the policy in a way employees can tolerate the rationale for it. John Webster, storage analyst for Illuminata in Nashua, New Hampshire says, "They need to understand that managing storage has nothing to do with the cost of the physical media, but with safeguarding their information and making it readily available to them without interruptions."
|
|
|
|
