As networked storage (e.g., SAN, NAS) matures by 2004/05, larger and more sophisticated implementations will drive the need for storage applications to become application and DBMS aware, which will expose immature storage-related security procedures. Through 2004, the value proposition and price emphasis will shift from hardware components to software and services, further separating and forming a storage operations discipline independent of DBMS, systems, and application network management.
As networked storage (e.g., SAN, NAS) and storage management (e.g., Fibre Channel fabric managers, storage resource management, SAN management) continue to mature through 2006, larger and more sophisticated implementations (moving away from self-integrated point products) will drive the need for functional storage applications to become application and DBMS aware, exposing immature storage-related security technologies and procedures. Through 2005/06, the value proposition and price emphasis will shift from hardware components to software and services, further separating and forming a storage operations discipline independent of DBMS, systems, and application network management.
IT organizations (ITOs) seeking to optimize both file-level and block-level (e.g., DBMS) access have had to trade off between best-of-breed solutions (e.g., EMC, HP, HDS for SANs, Network Appliance [NetApp] for NAS) and single-vendor solutions that are strong in one area and weak (or non-existent) in the other. Such tradeoffs result in either unoptimized performance or more complicated and perhaps redundant storage management tools (e.g., backup/recovery, data replication) in the weak area. In 2003, we expect leading vendors to significantly improve their offerings, with EMC enhancing its Celerra NAS, HP working to get into NAS with a credible offering based on EVA, and NetApp broadening support for its FAS900 SAN/NAS “filer.” HDS now offers NetApp filers as front ends to Lightning storage, which at least solves the issue of having a credible NAS offering. However, it does not solve all software-related issues (e.g., SnapVault versus XRC).
By 2004/05, we expect each of these leading vendors to provide the broad range of products necessary to qualify as enterprise-class suppliers for heterogeneous environments. Other significant vendors (e.g., IBM, Sun) will also enhance their storage product lines, but will continue to appeal primarily to their installed base of server customers. With neither an installed base of server customers to draw on nor a completely integrated SAN/NAS product line, HDS will diminish as a market force through 2006/07. Strong competition will continue to drive hardware prices downward by 35 per cent to 40 per cent annually through 2007/08.
The hype surrounding storage virtualization will finally bear fruit during 2003/04, with early policy-driven products being introduced by major vendors (e.g., EMC AutoIS, HP VersaStor/CASA, IBM Storage Volume Manager/StorageTank, Sun/Pirus, Veritas Enabled). However, as ITOs discover that applying some of these solutions requires currently implemented volume managers or file systems to be replaced on all affected systems with the storage vendor’s volume manager or file system, storage virtualization as the next-generation proprietary lock-in will become apparent. Although more choices will be enabled for underlying storage arrays, ITOs will become fully dependent on a single storage management vendor where brand switching can be quite costly. Thus, we reiterate our guidance for ITOs to choose such a vendor carefully and be prepared to live with the decision for five to seven years. Moreover, because changing file systems and volume managers can have significant implications for application deployment, we believe ITOs will make changes only as both server and application systems are upgraded. Given a typical three-year upgrade cycle, ITOs embarking on such a process now would not be finished before 2006/07.
Device management is a fundamental component of storage virtualization but a challenging area for vendors. Vendors must rely on either API swaps with other vendors, or the Storage Management Initiative Common Information Model (CIM). Application program interfaces (APIs) are problematic, because each vendor must negotiate the terms of the swap with each vendor individually. Moreover, vendors can pick and choose which features of their systems to expose through the API, and there is nothing to prevent one company from withholding or delaying future swaps for competitive reasons.
Sponsored by the Storage Networking Industry Association, the CIM model promises a uniform and widely available mechanism for communicating with storage devices. However, early CIM specifications are woefully inadequate (e.g., device discovery, configuration, and error reporting only). Moreover, CIM does not describe the behaviors a device will display under partial-failure conditions, which prevents vendors from providing service-level assurances. We believe the CIM specification will evolve slowly, and probably will not become robust before 2005/06. Until CIM is fully developed to describe the behaviors of all systems, a device may perform differently when managed by the native vendor rather than a “virtualizer.”
Storage management will remain composed of disparate products through 2005/06. Some vendors (e.g., CA, Tivoli) are proposing new storage “frameworks” (portals) to provide integrated storage management suites. Indeed, vendors are careful to avoid the “F” (frameworks) word, because they have earned reputations as time and money pits with little tangible return. For example, CA launched SANity twice but garnered little adoption despite free distribution. This was due, in part, to SANity’s positioning link to Unicenter TNG, which infected it with the framework perception.
To be successful, portals must strike a balance between lightweight implementation and feature bloat, yet do more than provide a simple, common launching point. ITOs should look for portals that provide a common information repository from which associated applications can draw data, yet be implemented and optimized separately. We believe such systems will evolve from major vendors through 2005/06, regardless of product name/category.
Business Impact: To minimize complexity and disruption, IT organizations should continue to pursue a strategy of consolidation by selecting one or two storage vendors.
Storage hardware systems are reaching a high degree of maturity, resulting in robust competition and choice. However, the practical implications of storage virtualization will be exposed during 2003/04, and IT organizations will realize that they must continue to select largely homogeneous solutions through 2005/06.