Although still in its infancy, grid storage promises to alleviate cost, flexibility, and reliability issues.
By Lynne VanArsdale and Warren Smith
Business-driven requirements for IT are propelling a new stage in the evolution of data storage-one that emphasizes “data” over “storage.” Networked storage enables more-efficient use of information assets. Yet there are end-user “pain points,” including cost, flexibility, and reliability, that must be addressed to enable the use of these new architectures in all contexts.
One of the new ways to solve these challenges is through grid computing, which involves the distributed execution of applications across dynamically available processors in a managed collection, or “grid,” of computers. This method of computing is highly efficient and addresses many of the resource allocation, cost, reliability, and flexibility issues that have plagued IT for years.
Storage and grids
Grid configurations can be significantly enhanced by specific storage architectures.
Grid storage is a relatively new term applied to storage architectures that provide data management via distributed peer-to-peer constructs. These new storage architectures support grid computing by enabling highly reliable, flexible provisioning of blocks-or files or other units of information-that can be requested using attributes from a pool of virtualized storage (see figure).
The ability to connect peer resources in a grid enables cost-efficient, flexible scaling when the peer nodes are robust enough to deliver the functionality required by the governing data management policies. Grid storage architectures improve flexibility by enabling the re-allocation of resources within the grid to balance the use of a variety of classes of storage to fit performance, capacity, and reliability requirements. And grids enhance reliability by enabling redundancy across peers, delivering robust fault tolerance via fail-over and non-interruptive recovery of components.
A few storage vendors have already announced grid storage technologies. For example, IBM Almaden Research has developed an architecture called Collective Intelligent Bricks (CIB), formerly known as “IceCube,” and HP Labs has demonstrated its Federated Array of Bricks (FAB) architecture. While these technologies are focused on disk-based block storage provisioning, HP has also recently announced its StorageWorks Grid, which allows for disk, tape, and other types of storage devices in a grid architecture.
The efficiency of grid storage architectures can be enhanced through the use of intelligent grid nodes, or bricks, that include value-add functions such as automatic configuration, self-healing, low-level data migration and replication, advanced diagnostics and predictive modeling, sophisticated caching, and command queue management. Also, it is possible to extend this functionality to allow applications to more cooperatively work with storage devices (e.g., by offloading content search and retrieval functions to storage nodes).
Intelligent bricks, or building blocks, will lower total cost of ownership. Although the grid may be managed to address intelligent brick functionality at a higher level, low-level support for this functionality simplifies system-wide (cross-brick) code and contains failure zones within each node, improving manageability. And consolidating functions within the bricks can reduce the software code necessary to make the grid function, reducing the probability of error. Brick hardware can also enable parts reduction to lower overall cost, complexity, and inventory requirements.
While grid storage technology is still in its infancy, a number of groups are considering the development of standards for grid storage to ensure interoperability and a high level of quality. For example, the Global Grid Forum is working on standards for grid architectures, and the Storage Networking Industry Association (SNIA) recently held a Birds-of-a-Feather session to drive grid standards in collaboration with other industry organizations.
Challenges currently facing these standards groups include grid security, policy-based management standards, common models and terminology, and seamless integration with data management software, especially in the areas of information life-cycle management (ILM) and content-aware storage (CAS). ILM encompasses the ability to provision storage based on attributes such as performance, availability, and security. CAS systems keep metadata about the stored data to enable quick and accurate management of the data, especially in the context of regulatory compliance. The SNIA Data Management Forum (DMF) is exploring the implications of these new data management technologies within the context of grid and virtualized storage.
Current grid efforts target high-end applications. However, future applications may include smaller bricks in smaller grids to meet the cost, flexibility, and reliability needs of SMBs. Object-based storage standards may enable high degrees of grid security, as well as embedded digital rights management and special archiving functions. The SNIA Object Storage Device Technical Work Group (TWG) is focused on driving object-based standards. And the SNIA’s Security Forum TWG is working on storage security technologies for grid architectures.
Grid storage holds promise for end users looking to improve cost, reliability, and flexibility of storage solutions. Intelligent brick components will enable efficient deployment of these new storage architectures, which will support advanced capabilities in performance, scalability, availability, security, and data management. The SNIA is dedicated to supporting these activities through its committees, work groups, and forums.
Lynne VanArsdale is a SNIA board member and senior director, strategic marketing, at Seagate. Warren Smith is a SNIA board member and senior manager of industry associations at Hewlett-Packard.