Intelligent Storage Networks

Storage needs a smart network

Intelligence at the network level is the next logical step in the storage networking evolution. It will provide users the ability to scale and consolidate their large SAN environments, and utilize previous investments more efficiently. by Arun Rawtani

Technological advancements and an insatiable demand for storage capacity continue to drive the rapid evolution of storage products and architectures. In fact, over the past several years, we have seen storage architecture transform from DAS to networked storage, creating the foundation for the next phase of the architectural evolution: Intelligent Storage Networks.

The need for intelligent storage networks is driven by members of the storage community who aim to ease the management, deployment and consolidation of small SAN islands into larger SANs. And many storage administrators want a unified way to replicate and seamlessly migrate information from one storage array to another—even if the arrays are from multiple manufacturers.

This need has created a great deal of hype and promise around intelligent storage networks, but savvy users of technology understand the importance of separating hype from reality, while they figure out how this next evolution of storage networking will play out.

EVOLUTION

Looking back at the DAS model, most storage intelligence was located at the server or the host. By the early 1990s, however, this model was struggling to meet increased demands for scalability, manageability, and business continuity. To address the emerging storage infrastructure needs, new capabilities were added to intelligent storage arrays, including greater consolidation and scalability through dramatic increases in storage performance, as well as advanced functionality, like local and remote data replication. Such technological advancements help drive the storage networking evolution by solving issues which the older architectures could not address.

Many companies today, have intelligence distributed at all levels of the storage hierarchy-in hosts (path management software, HBAs, and device drivers), storage networks, and storage devices. And now they want to optimize and leverage the placement of intelligence to maximize the business benefit. The way to do it is to look at the concept of intelligence at the network level. It is the next logical step in the storage networking evolution, because it will provide users with the ability to do something they could never do before—scale and consolidate their large SAN environments while utilizing previous investments more efficiently.

A wealth of proposed approaches has surfaced, but forward-thinking users have become increasingly skeptical about overly optimistic promises to take all the storage functionality that exists today and move it to the network. These approaches are sub-optimal because in many cases they are purporting to solve problems that have already been solved. What's more, many of them fail to address the key issues surrounding large SANs: scalability and consolidation.

DISTRIBUTING INTELLIGENCE

The proper guiding principal for distributing intelligence is threefold:

• The intelligence to manage and manipulate data at the Logical Unit Number (LUN) or volume level should reside close to the storage array.
• The intelligence to perform functions across the whole network should reside near network switching points.
• Application intelligence should reside in the server while simultaneously reaching out to all other points of intelligence in the network to ensure service levels.

As intelligence starts to move into the network, servers are becoming less aware of what's happening within the storage environment, allowing much more flexibility within the SAN infrastructure.

The business value of intelligence

With the relentless growth of data and increasing demands for application availability and performance at lower costs, customers aren't just asking for intelligent network applications. They want solutions to suit the scalability and consolidation needs of their networks. From a customer's perspective, the value resulting from moving the storage intelligence closer to what it controls can be broken down into three components:

• An extended choice of application, network and storage platforms at a lower cost, plus the ability for users to make decisions independently for better application and enhanced performance.
• Improved capabilities around storage delivery services that will provide dynamic utilization of network bandwidth and network capabilities.
• Reduced TCO via concentrated points of management—optimized storage and offloaded servers.

Free of the limitations and variations of storage hardware architectures, intelligent storage networks will deliver greater flexibility in storage management and data availability. Storage administrators, who understand the value of optimized, intelligent storage networks will be among the first to take command of previously complex storage infrastructures.

Intelligent network applications

Organizations are seeking to move away from application dependency to data independence. In order for this to happen, storage must become either application-transparent or accessible by any application according to hierarchical requirements. To help achieve this, two basic areas must be added into the network storage presentation. They are virtualization and data mobility.

Storage presentation refers to the way LUNs are presented to hosts and the ability to aggregate or split them. A great deal of industry discussion has been generated on this topic under the heading of virtualization. This capability can be accomplished today within a single storage array, but the network adds a new layer of virtualization that will span storage subsystems.

Network-based storage presentation services offer the ability to extend LUN presentation currently being accomplished within arrays and expand it across the entire fabric. This means that LUNs can be aggregated across different arrays to better utilize the storage capacity available in the infrastructure. In addition, network-based storage presentation makes storage pooling achievable, so that storage arrays can be virtually combined in a logical fashion and function as a single pool of storage.

Changing application requirements within companies and across industries mandate that storage be re-deployed in real-time. As a result, data mobility is required to insulate applications from storage location and data movement details. Data mobility can mask and automate the complexities of data migration (regardless of platform or legacy technology), and enable policy-based data movement. This matches the vision of many organizations that want to first segregate servers and storage and then interconnect them via multiple interoperating non-stop fabrics. Such a strategy forms the foundation of lifecycle data management.

Preparing for the future, today

Although much of this article is focused on future visions, steps can be taken today to ensure a smooth transition. To begin laying the groundwork for intelligent storage networking, storage administrators should partner with a company that has a proven track record for delivering complete solutions while they continue with their SAN deployments.

They should also continue implementing comprehensive management tools to both enable successful utilization of intelligent network capabilities, and ultimately, to ensure that data is automatically deployed to the lowest storage class needed to meet business requirements for data accessibility.

Arun Rawtani is a Country Manager for Technology Solution Group, based out of India for EMC Corporation