Focus server acquisitions on service value (e.g., obsolescence resistance, capacity on demand and workload management) and determine which applications lend themselves best to an outsource utility model rather than an ownership model. In planning server-based architectures, tailor deployments for the increased levels of computing diversity evolving by server role and placement. Server appliances will increasingly creep up the food chain to minimize costs and skills for limited-function applications and geographic dispersion. Avoid strategic investments in platforms with inconsistent track records, amorphous road maps, lagging market share, vendor financial instability and minimal outsourcing partnerships.

The Internet economy is disrupting the conventional server acquisition, evaluation and procurement process. Deployment options will escalate in complexity as users confront issues concerning mainframe-style consolidation vs. distributed and rack-mounted horizontal scaling, homogeneous vs. heterogeneous operating-system deployments and optimization, and emerging new value propositions, such as capacity on demand (COD).

Users will need new evaluation criteria to assess life cycles and obsolescence. The new thinking posits the following:

• Servers are instruments of utility,
• designed to deliver services over an infinitely expandable network,
• to a dynamically changing user population,
• tightly coupled to business partners via corporate data centers, Internet data centers and Web-hosting sites,
• functioning as service fulfillment vehicles for customer, trading, auction and consumer communities.

What will shape a new vision of server deployment? IS organizations must begin developing new disciplines in architecture planning and deployment, based on new concepts of server role and network placement. "Role" positions servers by function, such as back-end database platforms for online transaction processing (OLTP) applications, application servers, Web servers, network infrastructure and appliance servers. Greater economies and higher performance from commodity servers will drive deployments on the basis of server specialization, i.e., servers optimized for a specific function (e.g., network-attached-storage servers). "Placement" refers to architectural proximity, or position, in an end-to-end distributed system. We classify placement as being 1) inside the enterprise, 2) at the periphery, or edge, of the enterprise, or 3) outside, or external to, the enterprise but connected to enterprise systems.

The addition of the Web-enabled click streams to the legacy brick-and-mortar fabric will create a multiplier effect on server placements that will increasingly differentiate the value proposition by class of server. Scalability is a case in point. Users will need to achieve scaling both vertically and horizontally. Some applications, like large-scale OLTP or decision support and data warehouses, demand vertical scaling, achieved by symmetric or massively parallel processing. By contrast, server appliances will, by their limitations of footprint and high-density requirements, have no role here other than possible use as front-end application servers.

Many Internet-based applications will not really benefit from large-scale symmetric multiprocessing (SMP). Connecting tens or hundreds of servers with load balancing, clustering/failover and caching can provide effective alternatives. Thin, high-density servers will be less expensive (now dipping below $1,000) and quicker to deploy (in 15 minutes or so) — however, this will come at the cost of more management complexity and increasing footprint. Users wishing to build vertical functionality, such as an integrated middleware and application stack, and centrally manage and tune the system will continue to prefer the general-purpose approach.

Gartner's Application Server Evaluation Model (ASEM) provides a good starting point for users to set up an evaluation process of vendor platforms, technologies, relative market impact and opportunities for doing business (see Research Note COM-13-0535, "How to Get the Most out of Gartner's ASEM"). Users will find information on criteria and definitions for evaluating servers, the scoring process, the range of comparisons (e.g., by application and by platform technologies) and how to customize scoring for individual usage. The ASEM is an organic document that will be refreshed by Gartner on a regular basis to reflect the changing dynamic landscape of technologies, product announcements and functional evolution.

The ASEM can be used to perform an in-depth evaluation of the high-volume OLTP vendor segment (see Research Note COM-12-9270, "Gartner's ASEM: Update and Future Directions"). Gartner's evaluation scores and ranks the top contenders at the high end of the market, shows what changes have occurred since the previous versions of the model, and demonstrates how to make comparisons by different cuts or views. Users will recognize that no vendors have earned superior scores in all categories. The model will enable users to readily perceive the strengths and weaknesses of the vendor platforms, and it can be used to manipulate scores through changes in weighting, adding additional criteria as a form of what-if analysis. This process can be replicated for other classes of servers, such as those using Intel's technology (see Research Note COM-13-0118, "The OLTP ASEM Includes Focus on Intel Server Vendors").

The Server Value Proposition Will Shift From Point Solutions to Service Features

Successful next-generation servers will stress service features as part of a service-based architecture. Service processor features will routinely enable COD, quality of service, self-healing diagnostics and workload management. Self-healing is especially important to the ease of maintainability and reducing costs of servicing.

In the five-year planning period, server resources will be managed to a much more granular level, so that, no matter how much the workloads change second by second, from the unpredictable nature of Internet commerce, the server's intelligence will improve to dynamically allocate and deliver just enough computing to the user's specific policies and priorities. Vendors will contract with IS organizations to deliver computing and storage that protect and ensure performance and scalability and that can override CPU and disk failures and unexpected spikes in user service access, improving overall levels of availability.

An important precursor to this new "utility" model of computing is the COD concept. COD programs are being formulated by most major vendors to deliver servers with greater capacity than initial assumptions, but with ready-to-activate processor — and eventually memory, I/O and storage — resources to enable capacity expansion when situations should demand (see Research Note COM-12-8528, "Capacity on Demand: A Vendor Comparison," which describes in detail various vendor programs, how they differ, what their costs are and where the COD concept needs to evolve). Users will need to integrate these concepts and programs into future requests for proposals and consider their individual merits as part of the server selection criteria (described in our ASEM research).

Servers will have different design features, including: ease of use, setup and manageability; form factor; heat dissipation and power consumption; operating system; and price. While many products will appear to be commodity, the variety of functional uses for firewalls, cache, Web, storage, print and other functions will dictate a wide array of products and suppliers with different form factors, functionality and price. A key evaluation factor will be whether to choose a general-purpose server to standardize for procurement across a broad set of appliancelike applications, or to choose among specialized configurations and packages optimized to specific functions.

The changing market dynamics are propelling the use of function-specific servers, known as "server appliances" (see Research Note SPA-13-0155, "Server and Storage Appliances: Get Ready, It's Inevitable"). These deployments are keyed to network placement, ease of manageability, low price points and small form factors. They will undoubtedly move up the performance/function "food chain" with direct repercussions for users accustomed to selecting general-purpose servers on a broad enterprise scale.

Falling hardware prices will also enable vendors to “lend” users additional spare capacity to enable servers to handle unpredictable workloads as part of initial configurations. Initially, users will manually invoke the new resources and inform vendors. Eventually, service processor intelligence will monitor and detect extended CPU and system utilization beyond the initial set of parameters and automatically adjust capacity to just-in-time workload changes. Such metered usage will enable users to pay for performance delivered, rather than paying for spare, unused capacity.

Server Assets and Ownership: Outsourcing Evolves to Data Center Utilities

As server services increase in manageability and intelligence, users will be given more deployment options. These options will help drive improved investment protection. The resource can be owned outright, leased or accessed as a utility. Choosing asset ownership will require careful life cycle planning; choosing leasing requires familiarity with product and technology cycles. However, choosing a utility model shifts more of the burden of obsolescence to the utility supplier. The user will lock in service- and performance-level guarantees, and the utility will drive harder bargains with vendors as they increase their power and leverage in the market.

However, the increased benefits may also be more costly to administer and may create greater dependencies on the service supplier or vendors. Users should ensure that contracts specify that resources will be allocated by policy-based quality-of-service requirements to optimize response times and availability by applications and users. These new services and delivery mechanisms will also add complexities in terms and conditions and will demand a new user skill set to understand potential benefits and drawbacks, whether financial or in the conditions for compliance.

Server Improvements Extend the Concept of "-ities"

It has been common to talk about mission-critical application environments as sharing certain traits known as the "-ities" (e.g., scalability, availability, security and manageability). In the Internet economy, a new complementary set of "-ities" will emerge. As the server and storage utility model evolves, users must prepare for the next-generation "-ities" that will build on the original foundation. Here is a profile of "-ities" enhancements users should monitor:

Resource Fluidity: The ability of a utility supplier to create computing and storage architectures with a high degree of expansion, contraction and movement to satisfy diverse sets of user needs and workloads.

Workload Manageability: The ability of the resource utility owner to assign resources to applications on a demand and priority basis and to adjust the workload to changing demands by seasons, daily peaks and troughs, and upgrades or modifications in the application and business processes.

Leverageability: The economic advantages accrued by users in offloading the burdens of resource ownership, maintenance and management to a third party that amortizes costs on a pay-per-usage model.

License Flexibility: The options of users to pay only for computing power used or transaction and storage capacity consumed without forecasting resource requirements far into the future.

Multidimensional Scalability: The flexibility of evolving architectures to provide scaling from two perspectives: horizontal scaling (by small incremental, high-density, clustered computing resources) or vertical scaling through modular, in-cabinet performance enhancements such as cache coherent nonuniform memory access (ccNUMA), large-scale SMP and low-latency, switched CPU-memory bandwidths.

Fault Prevention and Recoverability: The ability to monitor workloads, to detect potential problems, bottlenecks and system weaknesses, and to provide means to restore data integrity, preserve work in process and protect in-flight transactions.

Service Quality: The ability of the user to request and the utility operator to deliver various levels of service guarantees to ensure business processes match the service levels demanded by the Internet economy in terms of response times and bandwidth.

Business Security: The ability of the utility operator to impose strict partitioning of resources among hundreds of service-level accounts, thereby ensuring that application or operational failures in one account are self-contained and isolated to prevent damage to other accounts.

Access Mobility: The ability of users to select and access computing resources by various modes of communications, including wireless, with options to use geographically dispersed resources so as to optimize routing and bandwidth costs.

Impact on Market Share and Competitive Positioning

The future of the server will evolve rapidly with technology and business innovation. As these changes occur, users will have to pay particular attention to the competitive dynamics, which could have significant impact on vendor market share and momentum. Vendors need to be aware of the market's dynamics and react quickly to adopt features, functions and services of next-generation server platforms to gain market share and competitive advantage.

The market's consolidation will continue during the five-year planning period. The primary operating-system environments will be Windows 2000, Unix and mainframe, along with vendor proprietary systems that have established market niches, such as the AS/400, OpenVMS, Compaq Computer's Himalaya and HP's MPE/iX. IBM, with the z900, and other mainframe vendors will guard the established markets for the mainframe — extreme scaling and availability in enterprise-critical environments. They will also target new-generation application areas, such as Web serving and enterprise resource planning, wherever mainframe attributes are valued.

RISC Unix vendors will focus on advancing further into the enterprise-critical space, and they will use Linux in an increasingly opportunistic or tactical move in infrastructure and integrated segments. Linux will evolve from formerly appearing as a “scattergram” of placements, toward becoming a Unix bastion against Windows 2000. As Linux support and tools (e.g., failover, management and load balancing) improve, Linux will increasingly compete with RISC Unix as well as Windows 2000. Windows NT/2000 vendors will concentrate on a three-part strategy: dominating the infrastructure and integrated segments, consolidating the existing hold on most of the high-volume generic segment, and capturing an increasing share of the enterprise-critical segment at the expense of Unix (with the advent of the Windows 2000 Data Center edition and more-scalable platforms, such as the Unisys Cellular Multiprocessing).

How well will GartnerGroup research forecast future events?