While he didn't go looking for a tiered data storage strategy for his storage area network (SAN), Tom Gonzales,...
senior network administrator at the Credit Union of Colorado in Denver, is glad he found one.
"Tiering was not on the requirements list we had for our storage area network, but it's definitely shown benefits, already such as making our system more stable, saving money on disks, helping us with compliance and allowing us to keep data around a lot longer," Gonzales said.
Tiered storage, which is also referred to as hierarchical storage management (HSM), enables storage managers to tailor their data storage purchases to the type of data they are storing. For instance, they can map frequently accessed data to high-performance, highly available Fibre Channel or iSCSI arrays. Storage managers can also map rarely accessed data to inexpensive, slower speed disks, or even tape libraries.
Avoiding a one-size-fits-all strategy is critical for SMBs that have high-demand storage needs, but don't want to overspend on their primary storage, according to Noemi Greyzdorf, senior research manager at IDC. "Tiering allows companies to reduce their data footprint and reduce costs by making sure that static or infrequently accessed information is moved off to lesser expensive storage," Greyzdorf said.
Tiered data storage offerings
In the past, storage managers who needed such a strategy had to jury-rig it themselves, manually moving data from one media to another. However, most data storage companies today, including Compellent, EMC Corp., F5 Networks Inc., Hewlett-Packard (HP) Co., Pillar Data Systems, Quantum Corp. and Sun Microsystems Inc., offer management tools that allow even the smallest firms to automate the categorization process.
According to Curtis Breville, senior analyst for storage management at Enterprise Management Associates consultancy, the goal for storage manufacturers is to match data to media based on its access frequency, performance needs or security requirements. However, each vendor's method for reaching this goal can be dramatically different. For instance:
- Compellent's Storage Center hardware-based offering moves blocks of data back and forth from tier to tier based on the frequency of access.
- Pillar's Axiom, which is also hardware-based, uses administrator-driven policies to move files based on the performance priority of the application.
- Quantum's software-based StorNext shuttles data to and from tiers based on user-defined policies and access requirements.
- Sun's software-based Storage Archive Manager-File System is also policy-based, but uses a combination of age, file size, file type and owner to determine the data's destination.
Breville adds that most solutions tap into Active Directory to ease policy setting and enforcement. "The biggest challenge for SMBs is nailing down the various tiers. Once you have those polices set, the rest is easy," he said.
IT teams should not to be derailed by the hefty upfront cost -- which can be upwards of $25,000 for software packages and $100,000 for hardware products -- because the benefits are tremendous. "In most cases, tiering is well worth the initial cost. Across the industry, 90% of data, once stored, is never accessed again. So therefore a SAN that addresses this by only having 10% of fast access will save a lot of money on primary storage," said Breville.
The Credit Union of Colorado's Gonzales found this out firsthand when he migrated from HP's Optical Jukebox archiving system to Compellent's Storage Center, a full-fledged SAN. The IT team had reached capacity on the 10-disk optical system, which held 20 GB per disk and was used to store member photos, scanned items such as licenses and signature cards, statements and other documents that required member or employee access. Gonzales also said that adding more disk space was going to be costly in terms of money and support.
"We're a small shop -- just over 200 employees -- and don't have a lot of expertise to manage and maintain our storage so the less we have to do manually, the better. With Compellent, tiering is built in and automated," Gonzales said. Rather than being stuck with a single-media strategy, he now boasts a three-tier approach.
Out of the box, he was able to set policies to map frequently accessed data, such as information on current members and Microsoft Exchange files, to 10K RPM and 7200 RPM Fibre Channel drives. Gonzales can also send more static data, such as former member account statements, to low-cost, slower speed SATA disks.
"We knew we were losing money by keeping the less-accessed data on the faster drives," said Gonzales. He estimated the Credit Union has saved $800 per disk because more than 95% of data overall could be moved to less expensive storage.
Breville said he's seeing this realization across industries. "The places where tiered storage works best are places that need a lot of data available for access. Hospital records, bank records, insurance claims -- you can keep all that data but have it sit on something far less expensive."
In addition, having a tiered strategy enables companies to alleviate quotas on e-mail. "With multiple tiers [behind the scenes], you can enable what is essentially bottomless email," he said.
While cost savings and space limitations are important to Woodbridge, Ont.-based environmental education site Earth Rangers.com, the fact that tiering reduces energy consumption is more critical, according to IT Systems Manager Rob Di Stefano.
Di Stefano uses Pillar's Axiom SAN to categorize data storage for the 50-person company. "While some of our applications such as Microsoft Exchange need to be highly available, others don't. So if you assign them to slower disks, they won't be spinning as much and will save energy, including power and cooling. If you're talking about 30 or 40 drives, that's a sizeable carbon footprint reduction," he said.
Setting a tiered data storage policy
Breville agrees and reiterates that all these savings come down to getting your policies right. Some sample policies include:
- No file over 2 MB will sit on tier 1 storage
- All .jpg or .ppt files automatically go to tier 2 or tier 3 storage
- All data owned by the CEO will remain on tier 1 storage
- Files will be moved to tier 2 storage after a month of not being accessed
Breville is not concerned that making people wait for stored data will negatively effect business. "If someone wants a bank record from four years ago, they will be willing to wait for 30 seconds for the file to come back from tape," he said.Sandra Gittlen is a freelance technology editor in the greater Boston area. She can be reached at firstname.lastname@example.org.