Who’s using fabric-based intelligence, and why?

By Heidi Biggar

Although surveys indicate that users would prefer to run storage services (virtualization, provisioning, volume management, replication, etc.) on fabric-based switches or specialized appliances or controllers, actual adoption of this technology has been slow.

While analysts believe users will begin implementing fabric-based intelligence in greater numbers over the next six months or so, widespread adoption is not expected for a year or more. At that point, analysts say users will be more familiar-and comfortable-with the concept of running storage services in the fabric and should have a more extensive set of technologies and vendors from which to choose.

In a research report, Nancy Hurley, senior analyst with the Enterprise Strategy Group (ESG), notes that a lack of products from larger vendors has contributed to the slow adoption of fabric-based applications. She also cites users’ lack of awareness of the technology and its benefits as another obstacle to adoption. (See “Intelligence in the fabric: Myth or reality?” Infostor, October 2004, p. 30.)

But as with most new technologies, a small group of users have been willing to become early adopters for a variety of reasons. For this article, Infostor spoke with three such users: two are running fabric-based applications in production environments, and one is currently in beta testing in a test and development environment.

According to a recent ESG survey, early adopters are already seeing significant benefits from running intelligent storage services in the fabric. a) 80% of users reported a hardware savings with a mean average of about 20% cost savings per year.
Click here to enlarge image

Although we contacted all the major switch vendors (including Brocade, Cisco, and McData), only three start-ups-Cloverleaf Communications, Maranti, and MaXXan-could put us in contact with users.

From conversations with these users, we discovered that while there were few similarities among their storage environments, they did share a common objective: to virtualize their storage environments so they could pool storage hardware from a mix of vendors.

b) “Simplified storage resource management” was the number-one-cited objective for moving intelligence into the fabric.
Click here to enlarge image

For example, Bezeq, Israel’s nationals telecommunications provider, implemented Cloverleaf’s iSN system as a way to virtualize its EMC storage environment so it could add lower-cost, non-EMC disk arrays.

Cloverleaf iSN is a network-resident system. However, unlike some competing technologies that provide alternatives to standard SAN switches, iSN incorporates switches from Brocade, Cisco, and McData. In addition to virtualization, iSN provides applications such as snapshots, replication, and backup.

c) Volume management, virtualization, and provisioning were the top-three services identified by early adopters and non-adopters for running storage services in the fabric.
Click here to enlarge image

“We had strong budget pressure to reduce storage costs, but we had no way of doing it if we only used disk systems from one vendor,” explains Yoshka Klein, manager of Bezeq’s IT division. By implementing Cloverleaf’s iSN system, Klein says he will be able to break the vendor lock-in that has limited his disk options over the past few years and ultimately reduce storage costs. He estimates that the company will potentially be able to reduce storage costs by about 30% in six months.

Klein currently manages about 300TB of data on EMC DMX and Symmetrix arrays. The 300TB are split between the company’s primary data center and an off-site disaster-recovery location. The company currently backs up to Clariion disk arrays and uses EMC’s SRDF software to replicate between its two sites.

Klein says that the company’s reliance on SRDF for replication precipitated the EMC lock-in: “It was the only vendor [at the time] that could provide this capability.”

As for other options, Klein talked to Cisco but it was unable at the time to provide the features he needed.

Bezeq began beta-testing the Cloverleaf system last month. Initially, Cloverleaf’s iSN will be used in non-production environments; however, Klein hopes to expand its use to production environments in about six months.

Take two

Like Bezeq, the need to pool, or virtualize, heterogeneous storage hardware, led the State of Kentucky to implement fabric-based intelligence. “Our backup provider thought it might be a good way for us to bring the different storage hardware that we owned together,” says Skip Hunt, IT manager. “We were running out of disk space, but we didn’t want to throw away what we already had.”

The State of Kentucky had multiple (three) SAN islands, a variety of disk storage systems, and capacity problems.

By implementing MaXXan’s MXV320 Intelligent Switch, Hunt was able to pool the mixed storage resources-which included an old Compaq 4100 disk array, a Hewlett-Packard MSA 1000, and an IBM FastT-from the three SANs and allocate storage as needed among servers and desktop users.

(Hunt subsequently upgraded from the MXV320 to an MXV500 and replaced the Compaq 4100 with an HP 4314 disk array.)

“We’re no longer concerned about disk space,” says Hunt. “The MXV500 allows us to expand the virtual pool infinitely.”

In addition to virtualization, the State of Kentucky is also implementing the MaXXan switch’s virtual tape library (VTL) capability to help it hit its backup window objectives. “VTL is a solution to a problem that everyone has,” says Hunt. “There is more data to back up than there is off-hour time do it.”

The State also plans to take advantage of the switch’s SAN IP and snapshot capabilities in the near future. Hunt says that the decision to do snapshots in the fabric, and not license the capability from IBM (an option on IBM’s FastT disk arrays), was a tradeoff in which cost and management were two of the deciding factors.

Hunt currently has 5TB to 6TB of disk capacity and about 15 Windows and Linux servers.

Take three

For Greg Graham, CEO of MetroDataPath, a managed services provider, the ability to provide data services to customers hinges on its ability to support heterogeneous storage arrays and to replicate data from its customers’ sites to its own facility, which includes Nexsan ATAboy2 and EMC Clariion disk arrays, as well as RAID subsystems from Apple.

MetroDataPath has a primary data site in Los Angeles and a secondary site in Las Vegas. Customer data is backed up over a gigabit optical network to the L.A. site.

“Every customer site we walk into has a different mix of storage arrays-JBODs, DAS, SAN, etc.,” says Graham. “We need to be able to create off-site mirrors from one vendor to the next.”

Graham says that by implementing Maranti’s CoreSTOR network-based storage controllers at his two locations (a CoreSTOR 3000 is installed in L.A. and a CoreSTOR 2000 in Las Vegas), he is able to create a single storage pool to service the varied needs (and disk resources) of MetroDataPath’s customers.

“The beauty of the CoreSTOR 3000 is that it allows a user [to not be locked into] a particular vendor,” says Graham.

The Maranti system supports both Fibre Channel and IP connectivity. Users that want to replicate data from their locations to MetroDataPath need to install a CoreSTOR network controller at their sites.

MetroDataPath implemented the two CoreSTOR controllers in October. In addition to Maranti, the company looked at competitive technologies from Cisco and FalconStor, but Graham says neither vendor’s platforms could provide the performance that MetroDataPath’s customers required.

This article was originally published on April 01, 2005