Networked storage systems break the boundaries.

Modular storage systems ushered in a new era in storage by delivering more flexibility and modularity than their monolithic storage systems predecessors. A new generation of storage system, the Networked Storage System, is now creating a dislocation in the marketplace by taking a page from distributed computing models and leveraging advances in storage networking.

Users with tight IT budgets have readily adopted modular storage systems as a path to acquire storage infrastructure at a fraction of the cost of large monolithic systems. As growth outpaced the extensibility (up until Fall 2003 most modular storage systems had a maximum raw storage capacity of 35TB) of these modular systems, users were forced to upgrade to a more powerful models or added additional systems. Upgrades required system downtime and resulted in business disruption. Increasing the number of systems meant an upgrade in SAN infrastructure and layers of software applications to manage theses systems. The complexity of managing multiple systems, inefficiencies in the utilization of this storage, and support costs quickly out-stripped the initial savings.

At the core of the issue is the architecture of a typical modular storage system. Most modular storage systems use single or dual storage managers housed in a common enclosure with some number of disk drives. Capacity is increased by daisy chaining additional disk cabinets. A fixed number of disk channels is an architectural limitation for capacity expansion. Host connectivity is limited by a fixed number of dedicated host channels. Performance is limited by the power of the pair of storage managers. Lastly, availability is disadvantaged by a dependency on a common backplane between the storage managers. Challenged to scale modular systems beyond the architectural limitations, vendors added virtualization software, storage resource management software, and a web of fabric switches.

[FIGURE 1 OMITTED]

Problem solved.

Not quite. Whatever happened to the simple elegance of modular storage and reducing cost?

The extensibility of modular mid-range storage systems is indeed limited by architecture; however, marketing also plays a role. The extensibility of modular systems is limited in part to protect the premiums associated with high-end systems, and the profits derived from the software and services required to create networks of modular systems.

The Challenge

Central to the challenge in deploying modular storage systems is balancing capacity, availability, and performance requirements versus budgetary considerations. These three fundamental dimensions of storage depicted in Figure 1 form the basis for storage system strategies. Aligning these needs with business and applications requirements is difficult enough; however; the situation is not static. Users have a seemingly inexhaustible appetite for additional capacity. Diverse and demanding application workloads create performance issues, while SLA's drive increasing requirements to improve availability. Current mid-range modular architectures fall short in delivering the flexibility to address the dimensions of this model in a simple, cost-effective manner.

Stretching the Dimensions of Modular Storage

Network-based modular storage systems deliver greater flexibility to address performance, capacity, and availability requirements. By eliminating the rigid dual-storage manager architecture found in traditional modular storage systems, networked storage systems provide linear scalability in performance and capacity. These storage systems employ a networked array of storage managers. Each storage manager yields additional connectivity to hosts, boosts data throughput, and increases system-wide storage capacity.

The Digi-Data STORM Xtreme Array, for example, has 24 Fibre Channel (FC) ports, a sustained data transfer rate to disk of over 900MB/s, and can support an array of up to 840 drives in a three-node cluster.

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

Simplified Management, More Capacity

From a management perspective, the array of storage managers acts as a single logical system. This management model greatly reduces the complexity of managing large installations while eliminating costly management software applications. Storage units within the array are pooled with all storage managers having visibility to all storage. Virtualization takes place within the array, affording a maximum in flexibility when isolating hosts and applications.

The virtualized storage pool may include both high-performance FC disk drives and economical Serial-ATA disk drives. In a networked architecture, expansion is no longer limited by the fixed channel connectivity of a pair of storage managers. Through dynamic expansion, storage capacity is increased without business disruption. In contrast to traditional modular systems, users are able to more effectively optimize the utilization storage within the system. High-capacity tiered storage supports an information lifecycle model where more information can be migrated from tape to disk to gain quick access to near-line content.

Multi-Tenancy

Multi-tenancy refers to storage systems capable of supporting both a diverse host operating system environment and diverse applications workload. Multiple host operating systems are effectively supported through the virtualization of the common storage pool. Applications diversity comes with the network of storage node architecture. In traditional modular storage systems, the dual storage managers are tightly coupled. The operational dependencies associated with this tight coupling reduce aggregate performance when applications have significantly different data characteristics such as streaming media versus email. Contrast this to the distributed architecture of networked storage systems, which creates a loosely coupled environment that eliminates the operational dependency and related performance hit.

Greater Availability With Geographically Distributed Nodes

The shared common backplane found in traditional modular storage systems creates an opportunity for complete system failure in the event of a catastrophic problem. The common backplane is a single point of failure. With networked storage systems, the n-way storage manager architecture not only does away with the common backplane, it creates an inter-manager correlation process to arbitrate individual storage node anomalies.

Unlike the captive storage manager found in traditional modular storage systems networked storage management nodes may be geographically distributed within the bounds of Fibre Channel distances. Systems may be deployed across a campus to protect against isolated site power failures, flood, or fire. This simple campus-wide disaster recovery solution is both inexpensive and easy to set up.

Networked Storage: The Solution for High Growth Applications

Unstructured corporate digital assets (or what is commonly referred to as reference information) presents IT management with a major challenge. According to the Enterprise Strategy Group, corporate reference information has a CAGR upwards of 76%. Reference information constitutes diverse data types and includes web content, CAD/CAM designs, presentations, rich media, scanned images, and the list goes on. IT management is faced with a need to develop a strategy to cost effectively support both the consolidation of storage infrastructure and growth of reference information through its lifecycle.

Network storage systems possess the flexibility through their modular architectures to address this multi-tenancy environment. Inherently pay-as-you-grow and void of the limitations of traditional modular systems, network storage systems provide a low total cost of ownership solution that scales in both performance and capacity.

Internet Service Providers, Applications Service Providers and Storage Service Providers are faced with a similar dilemma. Storage infrastructure must be quickly and easily adapted to respond to clients' unplanned growth requirements, changing SLAs, or peak demand scenarios. The profile of an xSP is in many ways similar to the multi-tenancy environment of corporate data centers managing reference information. Managing the cost of delivering a scalable and flexible services platform is a critical success factor for these companies. Networked storage solutions are cost effective and readily adapt to the dynamics of services businesses.

Bill Tomeo is president and CEO of Digi-Data Corp. (Columbia, MD)

www.digidata.com