Utility storage: Ready for a long haul?

Storage virtualization is a good first step toward full-out utility storage -- if you're willing to tough out rocky architectural, interoperability and management challenges

Utilization rates are up and spending down at early adopters. But storage virtualization remains tough to figure out, impossible to use in multivendor setting and hard to manage. Is it helping or hurting the utility storage vision?

First will come virtualization, then utility storage. That's long been the vision of how enterprise storage will evolve as IT grows increasingly dynamic and on-demand becomes business as usual.

The good news is that storage virtualization's day finally has arrived, the uptick in interest fueled by the success of server virtualization and reports of noteworthy results from early adopters. Pioneers cite big gains in storage utilization and decreases in device-level management headaches. They also share how they've avoided the spending sinkhole despite contending with ever-increasing data volumes. (Compare Storage Virtualization products.)

Cloud vs. utility: What’s the difference? Listen now.

"We've gone from 40% utilization of storage on the back end to up over 85%. Virtualization has saved us from going out and getting three times more disk than we needed, and we've realized a single point to manage connectivity between storage and the hosts," says Drew Kreisa, storage administrator at Mercury Marine, a recreational propulsion-engine maker in Fond du Lac, Wis.

Early adopters also have discovered, however, that the distance between storage virtualization and a utility-storage utopia is vast and full of what seem to be insurmountable challenges. Confusion - about the large number of architectural options, the lack of interoperability among different vendors' products and poor storage-resource management (SRM) tools for virtual environments - has muddied user expectations. As they wait for the industry to sort itself out, IT executives are left to relish the gains storage virtualization has brought them while pushing off their grand utility visions further into the future.

The wow of now

When Mercury Marine started looking at storage virtualization options five years ago, choices were limited. Not so today. The options for how to architect the abstraction of the physical layer are so abundant that IT executives should be wary, Kreisa says.

"There's a lot of confusion in the market. There are too many companies offering completely different ways to architect your network, which means you have to be careful not to bring a new component into your network that will blow away what you're already doing," he says.

In theory, storage virtualization lessens the complexity of managing, backing up, archiving and migrating data among pooled storage devices. With the technology, IT executives shouldn't have to get mired in details about the physical devices.

Picking the right approach to storage virtualization can be critical. First, a company must decide where it wants the storage virtualization to take place. For instance, it could choose a host-based system from such companies as Brocade Communications and Symantec. However, as these environments grow, they require their own operating system, host virtualization licenses, maintenance and software overhead.

— drew kreisa, storage administrator, Mercury Marine

A company also could deploy storage virtualization as part of the fabric with an appliance, such as IBM's SAN Volume Controller or with software that runs on the switch, such as EMC's Invista. Arun Taneja, founder of The Taneja Group consultancy, says the appliance-based approach is hot right now, while the switch-based approach doesn't have as much traction because of its higher cost.

A company that decides to go with a fabric-based strategy also must consider whether it's going to perform virtualization with in-band, out-of-band or split-path technology. In-band products, such as those from DataCore Software, FalconStor Software and IBM, allow both data and control information in the direct path of the host and controller. With out-of-band solutions, such as those from LSI, data flow is separated from control flow.

With split-path technology, which is employed by EMC, intelligent switches redirect control commands to external controllers and allow read-and-write I/Os to flow from the host to the appropriate physical storage array. Companies can virtualize their storage environments using an approach like that taken by Hitachi Data Systems, where controllers assign metadata to traffic as it flows from the switch into pools.

A final method involves the use of arrays. Such vendors as HP and 3PAR enable IT teams to create virtual logical-unit numbers (LUN) between the server and storage.

Fun at the block party

Yet another factor - whether a company wants to carry out virtualization at the block or at the file level - also is critical because, as Kreisa points out, there is little integration between the two. As part of a data-consolidation project, Mercury Marine deployed IBM's SAN Volume Controller appliance for block-level virtualization with Kreisa as the lead architect. Addressing the block level first let Mercury quickly alleviate its main stressors: difficulty in managing and expanding storage volumes, a lack of space on controllers, the need for a backup and restore process for expedient technology refreshes, and limited capacity for growth.

— Scott Christiansen, CSO, Leo A. Daly, explaining how his company has benefited from thin provisioning

Scott Christiansen, CSO at international architectural-engineering firm Leo A. Daly in Omaha, Neb., went in a different direction when he needed to consolidate data from the firm's 10 offices worldwide. As his network-attached-storage boxes expired, he replaced them with Dell's EqualLogic virtualized storage arrays that feed into Cisco Gigabit switches at each office. This allowed him to manage the storage pools as a single asset. Now, when users need space for their large AutoCAD and 3-D modeling files, he can tap into numerous resources.

Christiansen uses thin provisioning, a built-in feature of the Dell product, to automate resource allocation on his network. With thin provisioning, storage capacity is reserved only when applications write data to disk, ensuring maximum resource utilization at minimal cost.

"Traditionally, if an application needs 10GB of storage, a database administrator will ask for 40GB; the storage administrator will provision 100GB because he wants [the database administrator] out of his hair for a few months. That's 100GB earmarked for an application and won't be available to anyone else," consultant Taneja says. Thin provisioning lets IT executives such as Christiansen properly plan their capacity utilization.

Thin provisioning also saves money because it helps avoid the inherent power waste seen with today's storage, says Andrew Reichman, senior analyst at Forrester Research. "Normal provisioning leaves spindles spinning, consuming electricity and generating heat," he says.

Thanks to thin provisioning, Christiansen has been able to map data storage better and, by being able to focus on data as a whole vs. the individual store, has seen tremendous savings in technology cost and administrative time.

The pooled storage also bolsters fault tolerance, which in turn shores up disaster recovery and e-discovery plans, Christiansen says. "Storage is now one less thing that keeps me up at night. If anything goes down in one location, we can easily bring it up again in another site transparent to users," he says.

Virtual clarity With storage virtualization, IT executives have several layers of decisions to make, including what type of virtualization to create, where to deploy it and how to carry out that deployment. This Storage Networking Industry Association model shows how everything fits together.

Michael Schaffer, CTO at online book, music and video exchange Alibris in Emeryville, Calif., praises the fault tolerance he's seen from his managed 3PAR storage array, which is collocated with his network equipment at a facility in Sacramento.

Storage virtualization helps support the 5 million changes his global network of 10,000 vendors makes to his inventory database daily. The database's size and churn rate create a huge and dynamic demand for high-performance storage. "Our inventory database is the heart and soul of our business - the source of a competitive advantage as well as technical challenges," Schaffer says.

Today, Schaffer easily can allocate 100GB to one host or 1TB to another without worrying about the device-level details. "I do not know nor particularly care which drives are involved," he says.

The 3PAR system, which includes a mix of Advanced Technology Attachment and fast Fibre Channel drives, moves data automatically among storage tiers, Schaffer says. "Without downtime, I can move data from Fibre Channel to near-line storage or from a RAID 5 to a RAID 10 set" - and this happens without detailed planning and LUN carving, he adds.

Data deduplication, another popular storage-virtualization offshoot, is on Schaffer's wish list for its space-saving capabilities. Deduplication removes similar blocks of data and replaces them with hash marks, a process that offers such benefits as an increase in the time archives can be on disk, and better backup performance. The technology, however, requires a fundamental change in the way enterprise IT managers think about storage, Forrester's Reichman says. "For years, they have used multiple copies of data to ensure protection and availability. This is a case of the pendulum swinging the other way and reducing the physical copies to create a smaller footprint overall," he says.

Experts say enterprise IT managers could ease virtualization with a popular technology used with physical storage: snapshots. "Often overlooked, this is a good feature for making virtual backup easy, enabling migration, and for development teams, which can have an endless number of identical copies of production to data test new revisions against," says Mark Peters, analyst at Enterprise Strategies Group (ESG).

Now, the bad news

IT executives are finding success with single-vendor approaches to storage virtualization, but they admit that lack of vendor interoperability and other issues continue to stymie their attempts at utility storage.

Interoperability - or lack thereof - recently factored into Mercury's latest storage virtualization plans. The company plans on deploying file-system virtualization to reduce the number of storage entry points via servers it has around its network. Again, to avoid possible support issues, Mercury is leaning toward IBM's implementation of file virtualization technology from Network Appliance.

"Even if another product's features are similar, we have to consider our decision from an interoperability standpoint," Kreisa says.

This fait accompli surrounding interoperability has Ted Ritter, research analyst at Nemertes Research Group, pessimistic about the chances of storage becoming a utility soon. "If virtualization is limited because you're so closely tied to a single vendor, then you can't get to a utility utopia," he says.

Support for the basic Fibre Channel protocol doesn't mean much here. Managing storage devices and doing things like mirroring and moving copy aren't always compatible across different vendors, he adds.

Forrester's Reichman concurs. "Today, if you have one vendor's storage behind another vendor's virtualization console, it's very difficult to troubleshoot problems. A lot of finger-pointing ensues. This problem could be solved by making the virtualization and physical storage infrastructure more standardized," he says.

SRM, which provides the means of collecting information on heterogeneous resources, would benefit from standardization, too. SRM suffers from its inability to span virtualized environments from multiple vendors. This poses severe challenges for IT teams trying to troubleshoot performance issues. "You need deeper SRM integration to track application-performance problems back to specific spindle bottlenecks," ESG's Peters says. (Compare Storage Resource Management products.)

8 technologies for better data storage. View the slideshow.

SRM tools coupled with a virtualized environment create a roadblock to compliance, Nemertes' Ritter adds. When a company couples storage-resource and path-management tools with virtualization, "there is a lack of visibility into exactly what application data resides on which specific disk," he says.

This is a problem for organizations held to government and private-sector mandates. "If you can't tell where patient records are physically located [because they've been virtualized], then you're not [compliant with the Health Insurance Portability and Accountability Act]. The need to account for data is fundamental," he says.

The abstraction of the physical layer also affects disaster recovery and business continuity efforts. "When you remove the physical connection to where a given piece of data is and introduce virtualization that remaps the location of the data, breaks it up into components, fakes out the server, etc., you wind up relying on metadata and virtualization mechanisms to retain access to critical data," Forrester's Reichman says.

To counterbalance this, IT teams should develop policies about which data can be stored where and should use monitoring technology to generate virtualization-aware audits on that data, Reichman advises.

Thin provisioning can be a great benefit, but it can lull IT executives into a false sense of security. And this can result in disaster, Nemertes' Ritter warns. "There is a potential for underprovisioning your storage network to support system failures," he says.

1 2 Page 1
Page 1 of 2
SD-WAN buyers guide: Key questions to ask vendors (and yourself)