SANs require advanced RAID features

SANs require advanced RAID features

Techniques such as volume management can be particularly advantageous in heterogeneous storage area neworks.

Dave Simpson

Before you realize the benefits of storage area networks, you`ll have to rise to the challenges of configuring them with the right devices and software. Although a lot of the emphasis is currently on Fibre Channel host bus adapters, hubs, switches, and SAN management software, good old RAID is also a key player in SANs.

Some of the technologies that RAID vendors have been touting for years are just as important in SAN environments as they are in server-attached configurations. These include "dual-everything" designs (controllers, power supplies, fans, etc.), hot-swappable components, additional fault-tolerant features such as dual paths and automatic failover, data reliability features, heterogeneous server support, and remote monitoring software tools.

In addition, you`ll obviously need host-level Fibre Channel connectivity, which virtually all RAID, controller, and host bus adapter vendors have implemented. Some disk array vendors--such as Box Hill and Data General`s Clariion unit--are focused on "full" Fibre Channel implementations (i.e., Fibre Channel down to the drive level), while most other array vendors use Ultra or Ultra2 SCSI drives with Fibre Channel host connections.

But there are a number of technologies that go beyond those standard features. For example, one area to focus on can be generally categorized as volume management. It goes by as many names as there are vendors that offer it (which is few), but in all cases it provides data protection in heterogeneous host platform environments, and is particularly advantageous in Fibre Channel SAN environments.

At EMC, it`s called Volume Logix. Introduced late last year, Volume Logix was designed specifically for Fibre Channel storage networks with heterogeneous servers, including Windows NT, Unix, and MVS. The configura-tion management software resides on Symmetrix disk arrays, and allows storage administrators to control access to specific volumes for each server on the storage network. The benefit to end-users is secure disk access in shared-storage environments.

"With a SAN, you have a lot more servers connected to the same pool of storage, and you have to have a security and authorization mechanism," says Gil Press, director of market positioning at EMC. In contrast, with SCSI host interfaces, security was provided by the one-to-one host-to-array connections.

As with most EMC software, you`ll pay dearly for the added functionality: Volume Logix ranges from $15,000 to $35,000, depending on the size of the Symmetrix array.

Essentially, Volume Logix allows each host server attached to a Fibre Channel hub or switch to "see" only those volumes that that host should have access to. Without such data protection software, a server could gain unauthorized access to data "owned" by another host. Volume Logix is implemented in microcode in the disk array and as a user interface on a workstation or server. Administrators assign specific volumes to each host connected to the hub or switch.

Similarly, MTI Technology`s Volume Mapping software provides the same benefits by allowing storage administrators to partition and allocate storage resources to multiple hosts, including Windows NT and Unix (see diagram). Volume Mapping is implemented in controller firmware in MTI`s Gladiator disk arrays. Price: $15,000, regardless of the size of the array.

Volume Mapping was essential at U.S. Clearing, a financial services firm in New York that provides clearing and execution services. "We started with Unix only, and to add NT we had to have Volume Mapping for hard and soft addressing," says Mark West, vice president of infrastructure at U.S. Clearing. The company has 20 NT servers and three Unix servers in a cluster with three MTI Gladiator arrays with a total capacity of 3.4TB. Volume Mapping is used for data protection in a variety of applications at U.S. Clearing, including data warehousing, OLTP, and batch operations.

Packages such as Volume Logix and Volume Mapping are particularly beneficial in heterogeneous environments that include NT. "The dominant problem is NT," explains Kevin Liebl, vice president of marketing at MTI. "When an NT system boots, it grabs all the storage it can and automatically attaches to each volume it sees, even if they`re assigned to, say, a Unix server."

Analysts agree. "If you have NT in a SAN, volume management is critical because NT tries to suck up as much storage as possible and take command of it. This can be a real problem from a security standpoint," says James Staten, a storage analyst with Dataquest, a market research firm in San Jose.

Staten says the problem can be particularly acute for IT managers who want to use a SAN to consolidate the backup of multiple servers. "That`s the biggest problem because most companies are not building SANs to bring applications together; at this point they`re building them for backup purposes."

Volume management tools such as Volume Mapping and Volume Logix can be used in conjunction with other "levels" of SAN management software, such as the zoning capabilities provided by some Fibre Channel switch vendors. RAID-based volume management merely provides an additional level of data protection granularity.

A number of other high-end RAID vendors are about to introduce volume management techniques that will compete with technologies such as Volume Mapping and Volume Logix. For example, Hitachi Data Systems (HDS) plans to deliver the capability within the next quarter.

"Volume management is a key to SANs because you need to control access to the information on the drives, and you need to provide for alternate pathing," explains Dan Stromska, product marketing manager at HDS. "For example, companies want to make copies of databases for things like data mining, and you have to be able to easily manage that environment."

A number of other leading RAID vendors are working on volume management technologies that will provide similar end-user benefits.

For example, Clariion is developing "LUN masking" technology that`s due later this year, according to Peter Gibbs, director of marketing at Clariion. And Clariion`s OEMs will be quick to follow.

For example, Storage Technology plans to first introduce support for switching in its RAID arrays (within the next month) and then LUN masking (later this year). "Volume masking will provide the ability to share storage and data among multiple hosts, which is one of the main reasons users are excited about SANs," says Jim Hankins, a senior marketing specialist at STK.

The software approach

Another way to approach the volume management issue is from a software perspective. For example, Veritas Software`s Volume Manager provides volume management, but it doesn`t work with NT yet. Volume Manager currently runs on Solaris, and will support HP-UX next month, but support for NT is not due until mid-year, according to Mark Griffiths, director of software product management at Veritas. More advanced volume management is expected with the release of NT 5.0 (a.k.a. Windows 2000), because Microsoft licensed some of Veritas` Volume Manager code to replace the NT 4.0 disk administrator.

A potential advantage of the software-centric approach to volume management is hardware independence. In contrast, Volume Logix and Volume Mapping only work with arrays from EMC and MTI, respectively. Veritas is expected to announce within the next few months integration partnerships with a number of hardware RAID vendors.

Summing up the SAN volume management issue, Michael Peterson, president of Strategic Research, in Santa Barbara, CA, says "the idea is that in a SAN, arrays need an external administrative interface controlling volume access, security, and allocation--especially with NT, which doesn`t have a distributed file system." Peterson sees this class of software as the beginning of a hierarchy that progresses from homogeneous data sharing (shared access) to heterogeneous volume sharing and, finally, heterogeneous data sharing--the Holy Grail of SAN architectures.

In addition to evaluating volume management tools from RAID vendors, prospective buyers of SAN-based RAID should investigate third-party software packages that provide file sharing and volume management capabilities in heterogeneous environments. Sometimes (and somewhat erroneously) referred to as "SAN operating systems," these packages are available from vendors such as Mercury Computer Systems and Transoft, and provide volume management capabilities in ways that differ significantly from the approaches taken by hardware array vendors.

If you`re planning on putting RAID on a SAN, another feature to look for is fabric login, which will allow devices such as disk arrays to be automatically recognized in a switched fabric configuration and to take advantage of switch zoning. A standard for fabric login is due from the Storage Networking Industry Association as early as this spring. (SNIA is an industry organization that is working on a variety of SAN-related standards. For more information, visit www.snia.org.)

Trends to watch

Another trend in the RAID market is the increasing importance of software, a trend that will accelerate as the end-user community moves toward SAN architectures. Although performance and hardware architectures are still critical, software will increasingly be the arena where RAID vendors try to differentiate themselves. If you need more evidence, consider the fact that EMC CEO Mike Ruettgers recently stated that his company will spend more than $1 billion on software development over the next three years.

"Today, performance characteristics are much less a purchasing criteria for users or differentiators for RAID vendors. Even reliability is becoming more of a commodity," says Sean Derrington, a storage analyst with the Meta Group consulting firm, in Stamford, CT. "Software is becoming the focus, and will be a key differentiator."

A solid software strategy begins with array management software, which is available from any RAID vendor to manage its own arrays. The key trend in this space is toward Web browser-based tools that allow remote management and features such as temperature status monitoring, alarms, and paging. The ultimate goal, particularly in SAN environments, is centralized management of the large number of storage subsystems strung out on the storage network. For users, the main benefit of centralized management of distributed resources is a sharp reduction in cost of ownership.

At a higher level of software functionality, analysts say that EMC is leading the RAID pack with tools such as Volume Logix, Symmetrix Remote Data Facility (mirroring software for business continuity and disaster recovery), TimeFinder (software that eliminates scheduled outages by creating real-time copies of data), Data Manager (for direct, centralized backup), and PowerPath (load balancing and path failover software).

One interesting area to watch is what RAID vendors are doing to solve backup problems. Basically, the goal is to offload the primary applications network from backup-related bottlenecks, and to avoid going through the IP stack, to achieve "LAN-free" backup. The benefit to end-users is sharply reduced backup windows.

For example, EMC tackles this problem with direct-connect backup, moving data directly from Symmetrix arrays to tape libraries. But a variety of other RAID vendors are tackling the problem in other ways, often via partnerships with software vendors such as Computer Associates, Legato, and Veritas.

"Moving data at channel speeds with Fibre Channel gives you much better speed, connectivity, and distance capabilities," says Derrington. "That`s the next step in getting the data off the application network." Although largely a function of the backup software, these emerging models for high-speed backup will require intelligence embedded in disk arrays.

Another interesting RAID development is how--and how quickly--mainframe-oriented disk array vendors bring high-end functionality into their open systems arrays. Examples include Amdahl, EMC, HDS, IBM, StorageTek, and, via their VAX/VMS heritage, MTI and Compaq`s StorageWorks divi-sion (formerly Digital Equipment`s storage unit).

In part due to the trend toward SANs and network-attached storage, the RAID market is growing at a healthy clip. For example, disk array shipments rose almost 25% last year, to more than 1.3 million units, according to Disk/Trend Inc., a market research firm in Mountain View, CA. That translates into a more than $13 million market, a 13.4% increase over 1997 revenues.

Click here to enlarge image

Volume Mapping allows storage administrators to partition and allocate storage resources to multiple hosts to avoid data corruption.

Synapsys targets SANs with LUN masking

Synapsys Digital LLC, a relative startup in Palo Alto, CA, last month added LUN mapping capabilities in the 2.0 release of its TrueTriumph RAID array. Called BridgeAssist, the hardware-based mapping function allows administrators to interactively allocate logical partitions of the array to specific hosts, providing data protection in a storage area network (SAN) environment.

Like vendors such as Clariion and Box Hill, Synapsys is focused solely on RAID arrays with Fibre Channel at the disk-drive level. The company`s distribution strategy is based on a handful of resellers, as well as e-commerce (www.storenow.com). Synapsys also sells an NT-based MediaServer integrated with its TrueTriumph arrays, and provides systems integration services.

What do NT RAID users want?

The low end and mid-range of the RAID market (which still characterizes most Windows NT arrays), is careening toward a commodity market characterized by low end-user pricing, thin vendor margins, and little product differentiation. In most commodity markets, rock-bottom pricing is the key purchasing criterion. But not in the market for NT disk arrays (see figure).

According to a survey of 200 Fortune 1000 storage managers, conducted by Dataquest, a market research firm in San Jose, reliability and data integrity are the top concerns, with cost (both initial cost and cost to manage) ranking fairly low. A number of other features--most of which are de rigueur in the mainframe array market--also ranked high on the list, including disaster recovery and online backup.

Interestingly, cost to manage ranked higher than cost to purchase. That`s because NT puts significant management pressures on administrators. For example, Dataquest estimates that NT requires twice the amount of IT staff time than Unix, and eight times that of mainframes.

Click here to enlarge image

Data protection via virtualization

Disk array vendors such as EMC and MTI recently introduced data protection options for heterogeneous SAN environments. These firmware technologies allow administrators to carve up disk volumes and assign them to specific host server platforms to prevent data corruption. Storage Computer takes a different approach, which provides similar benefits, with its Virtual Storage Architecture (VSA).

VSA treats discrete disk arrays as a logical "storage fabric" that can be allocated and re-allocated to multiple hosts, including Unix and Windows NT servers. Storage Computer`s OmniRAID arrays allow administrators to set multiple data protection levels at the logical transaction or data set level. Hopping on the SAN bandwagon, the company recently began marketing its arrays as self-contained SANs.

While capitalizing on the SAN cachet, Storage Computer has been relatively slow to adopt Fibre Channel. "Fibre Channel is still developing from a protocol and reliability standpoint, and SCSI is proven," says John Webster, vice president of marketing. The company is still in the alpha stage of developing a Fibre Channel interface (based on Qlogic chip sets) that will be positioned primarily as a solution for connectivity distance issues.

In the mean time, Storage Computer plans to OEM Intelliguard Software`s Celestra technology for "LAN-free backup." (Intelliguard was recently acquired by Legato Systems.) A product introduction is expected within the next month or two.

This article was originally published on March 01, 1999