News Stay informed about the latest enterprise technology news and product updates.

What are clustered storage architectures?

Learn more about clustered storage architectures and the various vendor products on the market.

Clustered storage is appearing at all levels of the storage environment. The technology can overcome several obstacles storage professionals face today, including scale, performance, reliability, upgradeability and ease of administration.

For primary storage, there are two basic clustered storage architectures: tightly coupled clusters such as those offered by 3PAR Inc., and loosely coupled clustering offerings from companies such as Dell EqualLogic Inc. and LeftHand Networks Inc. (owned by Hewlett-Packard Co.).

Traditional storage offerings come in two flavors: monolithic systems like those offered by Hitachi Data Systems and EMC Corp.'s Symmetrix, or dual-controller modular systems from those same vendors and a host of others.

Traditional storage architectures require users to make a choice between a high-end, high-cost monolithic array that offers performance, scale and high availability, and a modular array that offers a lower upfront cost at the sacrifice of some level of performance, scale and reliability.

With a monolithic system, you're opting for increased capital expenditures upfront and over time for a reduced operating expenditure. With a modular system, you're accepting an increased operating expenditure for a reduced capital outlay.

Companies are forced to choose between monolithic and modular arrays and, in many cases, make compromises because their budget is constrained or they're not 100% sure of what their IO needs will be going forward. Enter clustered storage architectures. The goal of a clustered architecture is to offer the best of both worlds and effectively reduce both capital expenditures and operating expenditures. They fill the gaps for the small- and medium-sized business (SMB) and the enterprise that traditional storage architectures create.

Tightly coupled cluster architectures

Tightly coupled clusters tend to target the enterprise and service provider data center, as well as the gap created between modular and monolithic systems. They're the alternative to multiple modular systems and offer performance that scales right into monolithic territory by leveraging the ability to start small and grow cost-effectively as the applications demand.

These systems scale performance, IO and capacity independently. A tightly coupled clustered array typically has a proprietary physical backplane that controller nodes are connected into. While this backplane fixes the maximum size of the cluster, it delivers a very high-performance interconnect between nodes for load-balanced performance and maximum scalability as the cluster grows. Additional array controllers, IO ports and capacity are then connected into the cluster separately as demand dictates.

Similar to monolithic systems, the controller node backplane of a tightly clustered array supports cache-coherent controllers allowing port flexibility. Just like traditional, high-end arrays, any logical unit number (LUN) can be exported through any port on the array if that port becomes over subscribed or fails. Availability and manageability are high.

However, this architecture offers some meaningful differences. For example, it allows all of the resources in the cluster to be applied to any single volume. This parallel use of system resources opens a variety of possibilities. For example, it could enable the broader use of SATA storage for online applications, leveraging cluster technology and wide striping to gain high performance from drives traditionally used for nearline or archive applications. Load balancing and simplified management also result from this cache-coherent approach. Tightly coupled cluster technology simplifies storage management by presenting a single pool of capacity, controllers and connectivity resources that greatly reduce administration time while increasing performance and capacity utilization.

Loosely coupled cluster architectures

Loosely coupled architectures offer cost-effective building blocks that can start small and grow as applications demand. Typically, the interconnects between nodes are industry standard, often done over IP. This approach offers cost-effective configurations, but with less performance and scalability than the tightly coupled, controller-node backplane approach. Nevertheless, for an architecture aimed at SMBs, it's a worthwhile tradeoff and delivers management and scalability benefits vs. traditional modular arrays.

The nodes in these loosely coupled clusters offer performance, I/O and storage capacity within the same node; as a result, performance scales with capacity and vice versa. This approach provides very cost-effective, low-end configurations, but can become somewhat costly to scale to large configurations. Although the theoretical limitation of the size of a loosely coupled cluster can be quite high, scalability will be limited by the performance of the interconnect and the resulting tradeoffs in communication with the cluster controllers. Again, the architecture is well optimized for SMBs and smaller deployments, and provides a good alternative for companies seeking simplified management and a more flexible storage architecture.

Choosing the right architecture

For SMBs, management and growth of traditional modular arrays has proven costly. While loosely coupled, cluster storage architectures offer an effective alternative, modular systems (especially those with virtualization) offer a similar ease of use to that of clustered systems, while providing greater performance and connect flexibilities.

In the enterprise, the cost of monolithic arrays has forced organizations to seek out alternatives. However, multiple modular systems force management complexity and continue to lack the robust scalability and availability features of monolithic arrays. Tightly coupled clustered storage architectures shine in this area. They provide the performance, scalability and high availability an organization demands from a high-end enterprise array and make possible a wholesale shift of mission-critical data from monolithic to clustered architectures.

Dig Deeper on Data storage management