Problem solve Get help with specific problems with your technologies, process and projects.

SM2003: Seven steps for building and scaling storage networks

Bits & Bytes: Although many have a SAN, not everyone is content with them. Learn what GlassHouse Technologies' consultants suggested recently for dealing with SANs.

Are you happy with your storage area network?

That question was asked of one workshop audience at the recent Storage Management 2003 conference held in Chicago, Ill. Of the 150 storage users in attendance who had indicated that they'd already installed a moderate-sized SAN, this quick audience poll found that only half of them are currently happy with it.

GlassHouse Technologies' Richard Scannell, cofounder and vice president of strategy, along with senior storage consultant Bill Peldzus addressed common storage networking problems and concerns in their presentation on how to build a solid foundation for your SAN. The two offered seven steps, summarized in this article, to help you deal with ongoing storage networking complexities.

What to plan for when building a storage network
As networks, storage and databases intersect, new problems, processes and procedures arise that cannot be ignored. "You can't afford to not do something," said Scannell.

You need a balanced business approach that takes into account the technology, user requirements, organization and cost. Many vendors will only talk to you about technology and cost, said Scannell. You have to also keep in mind your regulatory and business requirements to have an effective storage network.

In addition, interoperability continues to be an important issue. "If we're not even talking about products, there are over 67,000 (configuration) combinations," said Peldzus, referring to the multitude of settings you could make with HBAs, bridges, routers, switches, microcode, firmware levels, etc. "What you have probably has not been tested and if you're going to put anything new in -- test, test, test. Test it with the mindset to break it," he emphasized.

Budgets, ROI and TCO
"In 2003, you can spend -- but you can only spend to save," said Peldzus.

To help you justify a SAN investment, you need to determine the value of your data and if that value will decrease over time. Then, you should appropriately congregate data and rationalize storage requirements as that data moves up or down the value stream, according to Scannell. This process means understanding what you have and how it's analyzed. Scannell suggested doing so by creating classes of storage based on architectural attributes, then tracking the TCO, utilization and occupancy for each class.

Non-technical considerations
"Management is moving from a technology operation to a strategic element of business performance and you're going to get dragged in whether you like it or not," said Scannell.

To adapt to this change, you need to adjust storage governance, or the exercise of authority over storage and the performance of functions, according to Scannell and Peldzus. For instance, rather than guesstimating your storage needs, you need to predict them. Rather than just reporting utilization, you need to know the occupancy of your storage devices. Make benchmarks into operating standards. Turn tactical decisions into strategic ones, they said.

Also, before you buy, decide what your RFP will look like. "Vendors are good at making RFP responses so confusing it can't compare to anything else," according to Peldzus. Ask for a line item RFP.

Storage classes, decisions and their effect on applications
You should create three classes of storage -- low, medium and high importance -- and determine which are mission critical, according to Scannell. The applications that are ranked the highest should be on the most valuable system. "Think of it as building a house. You build a house to protect your family, but you buy a safe to put in the house to protect the family jewels," said Scannell.

Scannell suggested a number of ways to improve performance of applications based on the storage architecture you select. For instance, you can manage more gigabytes per administrator in a pooled environment, achieve greater availability by implementing a three-node cluster using SCSI or improve disaster recovery with an architecture that doesn't involve the server for data replication. For data protection, he recommended avoiding the purchase of another tape drive for every new server. "Having tape drives on each application server doesn't make sense in large environments," he said.

Also, keep in mind multiple tape drive configuration issues and hazards. For instance, without persistent binding, the physical-to-logical mapping of tape drives can be reordered every time there is a reboot, according to Peldzus. To prevent this problem, replace drives before you reboot and confirm mappings before you begin backups, he said.

Tips for migrating data onto network storage
Scannell and Peldzus told the audience that there are multiple approaches to data migration. There's the simple approach (backup tapes), the moderately difficult approach (mirrored copy, snapshots, etc.) or the complex approach (specialized data migration and virtualization). No matter which approach you choose, they stressed the need to prepare thoroughly for a migration.

Your migration should occur in three phases. First you have to introduce storage networking with high availability into the environment, then you have to move from direct-attached storage to the SAN and determine which data you want to migrate first according to its value. Finally, you need to ensure that the SAN is enterprise ready.

"If you're planning a migration over a long weekend, know what time to "turn back" if it's not going as planned...whether it's hour 25 or hour 26," said Peldzus. Don't forget to factor import/export times into databases or applications.

What to do after it's up and running
Once your SAN is installed, turn your attention to lifecycle storage management and troubleshooting, Scannell said. (Lifecycle management involves demand planning and provisioning to address customer needs and vice versa.) Here, you need to align value with cost and make sure the most valuable data is always available.

For effective troubleshooting, Scannell and Peldzus said you should be able to answer, "Yes," to the following questions:

  • Have you monitored for normal?
  • Is your staff trained to work with the technology?
  • Are you prepared for "what if" scenarios?
  • Are maintenance contracts consistent?

Planning for the future
Finally, as you plan for the next quarter, year and beyond, various technologies will come into consideration for your storage network. These include IP storage, disk versus tape for backup, InfiniBand and storage management products, according to Scannell. But it's important to "look before you leap," he said. Some words of caution: Be careful of baseline requirements, what is actually supported by the technologies and outrageous vendor claims.

For more information

Tip: Scale your SAN for tomorrow

Bits & Bytes: Building a SAN from A to Z

Expert advice: DR considerations when building a SAN

Dig Deeper on Data storage strategy

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.