To realize the cost savings and agility benefits of a storage-area network, infrastructure developers should focus on enterprise storage principles (i.e., decoupling storage and server life cycles, creating shared storage services, centralizing and consolidating storage, offloading storage traffic to a storage-specific network).
Though 2001, maturing electronic business initiatives will increase enterprise demands for information storage. To reliably deliver on these initiatives, IT organizations must develop a robust, agile, and economical storage infrastructure.
Historically, organizations have purchased (and deployed) storage resources (e.g., mechanical disk, backup/recovery [B/R] software, tape devices, robotic libraries) on an as-needed and individual application/server basis. Through 2001, many organizations will realize storage is another critical infrastructure (comparable to network and security infrastructure) that must be developed to support rapidly changing business requirements.
By 2004/05, 80% of organizations will have defined new enterprise storage strategies based on storage-area networks (SANs) and network-attached storage. Most organizations will require both technologies, with SAN accounting for 80% of the storage market by 2003. Hardware and software storage consolidation and improved staff efficiency will be the primary business drivers. The common theme is to improve storage management practices by using enabling hardware and software that reduce the operational impact of increased storage capacities.
META Group has long asserted SANs are a set of principles, not a technology. To create an adaptive storage infrastructure that realizes agility and cost-reduction benefits, infrastructure developers must embrace these enterprise storage principles and select enabling technologies consistent with them. Four enterprise storage principles must be observed.
Decoupling Storage and Server Life Cycles.
Storage is the dominant cost driver in server purchases (more than 70% of the server purchase price by 2002). To reduce cost and increase flexibility, IT organizations should decouple storage and server life cycles (see Figure 1 in Addendum). This enables the storage and server life cycles to be managed separately, according to their distinct technology refresh rates and business requirements. By 2004/05, 75% of new storage deployments will be decoupled from the server.
Through economies of scale derived from decoupling storage selection from the server, IT purchasers will be able to negotiate volume discounts and, just as important, standardized discounted maintenance contracts. This establishes consistent benchmark storage costs across servers from various vendors.
Greater flexibility will also result because storage can be scaled independently of the server’s physical capacity. Moreover, the server can be scaled (e.g., upgraded, swapped out) without replacing the associated storage, minimizing the associated data migration issues. IT organizations must recognize storage as an infrastructure asset that is invested in and optimized to provide the greatest value to the business.
Creating Shared Storage Services.
The greatest benefits of a SAN are derived from packaging storage as a shared service and adopting the practice of operations excellence. META Groups research indicates storage management costs over a three-year period (including B/R) vary from 4x-15x of the initial hardware purchase cost; organizations deploying shared storage services are closer to 4x.
IT organizations should create a center of excellence (COE) for storage and develop standard storage processes (e.g., B/R, disaster recovery [DR], on-demand storage, performance monitoring/management) that can be applied across the existing server technology silos (e.g., Unix and NT). The application of common procedures across server platforms ensures standard service levels in crucial (and often neglected) disciplines such as B/R and DR.
The data and media COE should create a menu of storage services that are matched to a range of the business’s requirements, such as standard managed storage (e.g., limited redundancy and standard B/R) or mission-critical storage (e.g., remote mirroring, online snapshot image of database, remote tape vaulting). Servers subscribe to appropriate storage services, reaping the benefits of incremental storage processes improvements (e.g., tape automation, remote vaulting, or LAN-free backup).
By adopting operational excellence in storage, the total cost of ownership can be driven down while improving or holding constant the service quality (determined by the business). Moreover, a shared storage services model complements the application subscription chargeback method adopted by leading IT organizations.
Centralizing and Consolidating Storage.
Consolidation drives down storage costs by increasing resource utilization, simplifying storage management and enabling storage administrators to manage a larger data capacity. Consolidated storage requires less excess capacity (due to a larger margin of error in capacity planning) than distributed storage because capacity can be dynamically allocated/reallocated as needed from one large pool.
Consolidation simplifies the environment by reducing the number of storage-related products (storage management tools, B/R tools) that staff members must become skilled in. Consolidation should also be extended to cover shared tape libraries. The use of common management tools reduces the largest operational cost by enabling storage administrators to manage significantly more data (2x-3x) and considerably reducing the labor required for tape management.
The three essential features required for storage consolidation are partitioning, workload management (WLM), and incremental scalability. Partitioning ensures a server has access only to the resources it is allocated. WLM enforces service levels by guaranteeing the availability of minimum levels of critical resources (e.g., network bandwidth, controller cache). Incremental scalability enables the capacity of the storage pool to be increased to meet application demands without overprovisioning initially or taking all systems down to increase the capacity.
Offloading Data Traffic From the Application Network to a Storage-Specific Network.
This principle enables connectivity to be optimized for storage requirements. For example, in online transaction processing environments, throughput of random reads/writes and data integrity (from server to disk storage) favor a switched Fibre Channel network that can be scaled and dedicated for server/storage I/O requests.
By providing any-to-any connection, the storage network enables disks to be optimally allocated to satisfy dynamic changes in storage capacity. A dedicated storage network ensures adequate bandwidth is allocated to time-sensitive operations such as recovery without affecting the existing running applications. //
This is an abbreviated version of META Group’s report.
Copyright 2001 META Group Inc.
208 Harbor Drive, P.O. Box 1200061, Stamford, CT 06912-0061.
Web: http://www.metagroup.com. Telephone: (203) 973-6700. Fax: (203) 359-8066.
This publication may not be reproduced in any form or by any electronic or mechanical means, including information storage and retrieval systems, without prior written permission.
All rights reserved. Reprints are available.