Modular RAID arrays are taking over monolithic arrays in applications such as storage consolidation, enhanced backup, and disaster recovery.
By Tim Rasmussen
RAID systems were developed more than a decade ago to solve the common problem of failed disk drives causing applications to fail and end users to lose access to their data. Modular RAID systems were direct-attached to open systems platforms (Unix, Windows, etc.), and although they provided a high level of availability, they were not particularly high in performance or easy to manage. Considered "midrange" systems, they did not have the functionality necessary to support enterprise environments.
They were most often deployed in small to medium-sized companies or at the departmental level of larger companies. Enterprise environments, meanwhile, were dominated by monolithic arrays with more-robust features and higher quality of service capabilities.
The shift toward modular
Today's modular disk systems, however, have greater functionality, performance, capacity, availability, and heterogeneous server connectivity options and are being deployed in higher-end storage environments. They are no longer excluded from supporting mission-critical applications inside the data center and, in many situations, can be successfully implemented in the enterprise space. As a result, the naming convention of "midrange" and "enterprise" RAID systems no longer applies. Now the industry has adopted the terms "modular" and "monolithic" to describe the two classes of storage devices.
Modular RAID arrays now offer most of the software functionality available in monolithic storage systems (and at much lower prices), including
- Sophisticated management software for storage provisioning and monitoring;
- Snapshot software for point-in-time copies;
- Remote replication software for disaster recovery;
- LUN masking software to map storage capacity to specific servers; and
- Alternate path software for higher availability.
The advances in quality of service capabilities and software functionality enable modular disk systems to be used in a number of solutions such as the following:
A modular disk array's features—scalable capacity, high performance, high availability, ease of management, and cost effectiveness—lend themselves to storage consolidation projects. In addition, modular arrays offer a compact footprint for optimum capacity per square foot.
Storage consolidation using a modular disk array provides many of the same benefits as using a monolithic RAID system: increased performance and data accessibility, greater storage utilization, and more-efficient storage management. Modular disk is an ideal choice when consolidating storage in open systems server environments where a few to a couple dozen servers are connected. As more servers are added, modular storage devices can be added to a storage area network (SAN) and managed from common interfaces (if the arrays are from the same manufacturer). Conversely, monolithic storage systems—with the ability to scale to more than 100TB and support hundreds of servers simultaneously—are typically still the preferred choice in the largest, most mission-critical environments.
Modular storage systems now provide enhanced backup options through point-in-time snapshot capability, which allows a copy of a logical unit number (LUN) to be made internally to the RAID system. Two types of snapshots exist: image copy and view copy.
An image copy snapshot is a full block-level mirror of the primary LUN. The image copy takes up an identical amount of capacity as the primary LUN and can be reassigned to other servers on the SAN and used for testing, data mining, or data warehousing applications.
A view copy snapshot replicates just the pointers to the data on the primary LUN. In the event of a write operation from the server, the RAID system performs a copy-on-write. That is, the existing data to be updated is moved to the snapshot copy before the primary LUN is updated. Because only updated data is copied to the view image snapshot, relatively little capacity is required. View-copy snapshots are ideal for enhancing backup environments. The view copy snapshot can be redirected to a dedicated backup server, providing an off-host backup with zero impact to the production server.
Until recently, the ability to perform storage-to-storage replication was limited to monolithic storage systems. However, with remote replication software, modular systems can now mirror data to a physically separate RAID system ranging from short distances inside a building to long distances over a WAN. With this capability, modular storage systems can now be used as a key component in high-availability disaster-recovery applications. Replication and snapshot technology can be used together in disaster-recovery solutions that also enable off-site backups, further enhancing an overall data-protection strategy.
Many vendors manufacture arrays that can be categorized as modular disk systems. Each offers high-performance processors, cache to enhance performance, and small shelves installed in a standard 19-inch rack. The main differences are with software functionality, product maturity, and host connectivity options.
During the next few years, the capabilities of modular disk arrays will be further improved and these subsystems will continue to be used in higher-end storage environments.
According to Gartner Dataquest, in 2000, the average amount of storage managed by a single modular RAID array ranged from 240GB on the low-end to 1.2TB on the high-end.
Gartner predicts that, by 2005, average capacities on the high-end will increase nine-fold to 10.8TB.
Similarly, larger monolithic RAID arrays in 2000 managed an average of 1.6TB on the low-end to 4.2TB on the high-end. Gartner estimates that, by 2005, average capacities on the high-end for monolithic devices will increase to 38TB per system. Given these predictions, it's likely there will always be a market for monolithic devices; the line where modular ends and monolithic begins will continue to change, however.
Tim Rasmussen is a senior product manager at Datalink (www.datalink.com) in Chanhassen, MN.
Modular vs. monolithic: What's the difference?
Monolithic storage systems were initially developed for mainframes and eventually were enhanced to support open systems platforms. The three dominant players in this market are EMC, Hitachi Data Systems, and IBM.
Monolithic systems support 24x7 mission-critical applications by providing the highest levels of availability, performance, and functionality. The monolithic architecture uses several large frames, each roughly the size of a refrigerator, housing everything from host interfaces, large amounts of cache, numerous RAID controllers, and hundreds of disk drives.
Modular storage, as the name suggests, uses building blocks of controller and disk drive shelves allowing very customized and scalable solutions. The controller module contains host interfaces, cache, dual RAID processors, and interfaces to the disk drives. The drive modules are simply JBOD (just a bunch of disks), typically housing up to 15 disk drives per shelf. Modular storage systems are installed into standard 19-inch racks and, with a shelf size of 3U (5.25 inches), can support up to 240 disk drives while taking up relatively little floor space.
The modular architecture offers a "build as you grow" approach in that as more capacity is needed, additional drive shelves can be added non-disruptively. Likewise, as servers are scaled out, more modular disk systems can be added if designed into a storage area network (SAN).
Targeted specifically at open systems servers, modular disk arrays can scale anywhere from a couple hundred gigabytes to more than 30TB in capacity. With high-performance processors, sophisticated caching algorithms, and many tuning parameters, modular arrays can be configured for either extremely high I/O applications, such as Oracle or SQL Server database environments, or for high-bandwidth applications such as medical imaging or CAD systems.