...............................................................................................

Cahners In-Stat Group White Paper

April 25,1997

Table of Contents

• Summary
• Distributed Computing Drives a NewServer Model
• Windows NT Server Provides Lower Cost of Ownership
• Demographics and Survey Methodology
• Cost-of-Ownership Methodology
• Initial Costs
• Installation Costs
• Outside Services
• Upgrade Costs
• Total Initial Costs
• Ongoing Costs
• Ongoing Server Management Costs
• Ongoing Applications-Related Costs
• Total Ongoing Costs
• Total Cost of Ownership
• Lower Cost of Ownership, Higher Satisfaction
• Conclusions
• Total Cost-of-Ownership Model

Summary

Cost of ownership is a critical concern for most companies investing in distributed computing. This study, based on interviews with 500 respondents who manage either Windows NT Server or NetWare / UNIX server operating systems environments, demonstrates that Windows NT Server costs 19% less to operate on an annual basis. This total cost-of-ownership analysis includes initial systems costs, software costs, installation costs, services, upgrades, ongoing management costs, and applications-related costs.

The objective of this study was to analyze the total cost of ownership of these two server operating environments:

• Novell NetWare used as a network operating system coupled with UNIX used as an applications server environment, and
• Microsoft Windows NT Server used as both a network operating system and an applications server environment,

in order to assess whether Windows NT Server’s delivery of a single environment for networking and applications results in a cost advantage for its users. The results of the study emphatically indicate that Windows NT Server does indeed deliver lower total cost of ownership than Novell NetWare used in conjunction with UNIX. In total, the cost of ownership for Windows NT Server environments is 19% less than NetWare / UNIX environments.

Survey respondents also indicate that cost of ownership is a critical issue for them.

The study, commissioned by Microsoft, was conducted via telephone interviews in the first quarter of 1997 among 500 IS professionals in the United States. The names used for these interviews were derived from independent industry databases. Respondents met the following criteria:

• They were qualified as the person at their site most knowledgeable about server operating systems management and development.

• Their organizations were running either Windows NT Server or a combination of NetWare and UNIX on 70% or more of their servers.

• They were using either Windows Server as both their network operating system and their applications server operating system or NetWare as their network operating system and UNIX as their applications server operating system.

Fifty respondents were included in the study from each environment for each of the following size groups:

• 1 to 25 users,
  
• 26 to 50 users,
  
• 51 to 499 users,
  
• 500 to 999 users, and
  
• 1,000 to 3,000 users.

Cahners In-Stat Group’s cost-of-ownership model uses interviews with technology users to determine:

• initial systems costs,
  
• labor time associated with initial systems implementation,
  
• out-of-pocket costs associated with implementation and training services,
  
• out-of-pocket costs associated with upgrades required to run the systems,
  
• labor time associated with upgrades required to run the systems,
  
• labor time associated with ongoing use and management, and
  
• appropriate time frame for considering return on investment.

Cahners In-Stat Group then uses industry average salary information to convert labor time to labor costs in order to derive a total cost of ownership. In order to control for variations in size of environment across different products, Cahners In-Stat Group calculates cost of ownership for operating systems on a per-user, per-site, and per-server basis.

In order to ensure that all labor time is accounted for, Cahners In-Stat Group breaks the labor time out into meaningful subcategories for each greater labor category: initial systems implementation, upgrades, and ongoing use and management. These subcategories are derived from Cahners In-Stat Group syndicated survey and focus group research on related topics.

Arithmetic means for each element of the cost of ownership for each size group are entered into the total cost-of-ownership model. Each mean is tested for statistical significance and outliers are eliminated. Seventy-six percent of the means tested as significant using t-tests. According to the t-tests, the most significant differences occurred in the less than 50 user categories and the 1,000 to 3,000 user categories.

Initial Systems Costs

The initial systems costs is a key element of the total cost of ownership. For many users, a product advertising low cost of ownership is a red flag for a product that has a high purchase price. In an environment of ongoing pressure on information technology budgets and decreasing product life cycles, high-cost products are difficult to justify; there simply is not enough time in their useful life to achieve payback.

The initial systems cost for the purpose of this analysis includes:

• the initial purchase price of the operating system,
  
• the initial purchase price of the server hardware,
  
• the initial purchase price of the database,
  
• the initial purchase price of applications development tools, and
  
• the initial purchase price of packaged applications.

The purchase price for the operating system and hardware were gathered on a per-server basis and multiplied by the number of servers to arrive at a total initial outlay for the base system.

In total, Windows NT Server users spent 25% less than NetWare / UNIX users on their systems.

The cost advantage is greatest for Windows NT Server in the 1 to 25 user and 26 to 50 user size classes ¾ over 60%; the advantage drops to 25% in the 51 to 499 user size class. In the 500 to 999 user size class, Windows NT Server has a disadvantage; NetWare users report a price drop in this size class that may be due to discounting for specific industries or channels. In the 1,000 to 3,000 user size class, Windows NT Server has a moderate initial cost advantage over NetWare / UNIX environments. The per-server cost for operating systems and hardware is 20% lower for 1,000 to 3,000 user sites, but the total cost is only 15% lower. The value-added software cost is insignificantly higher for Windows NT Server users for this size class as well.

The labor associated with initial installation of a system is the next factor examined in the cost-of-ownership model. While it is another one-time cost, it is an important consideration for growing companies that expect to add new servers to their environment. A rapid installation can facilitate responsiveness to growing end-user demands.

In examining the labor cost associated with initial installation of a new server operating systems environment, Cahners In-Stat Group includes the following labor categories:

• hardware setup,
  
• installation of the operating system,
  
• installation of the database,
  
• installation of development tools,
  
• installation of packaged applications,
  
• initial applications development, and
  
• IT staff training.

Across both environments, the most time-consuming aspects of initial setup are applications development, training, and the initial setup of the hardware.

Overall, Microsoft environments took less time for initial installation and applications development than NetWare / UNIX environments, resulting in an overall 37% lower labor cost.

Windows NT Server requires less labor overall to install; cost advantages for installation are greatest in the 500 to 999 user size class. Windows NT Server respondents spent significantly less on training and value-added software installation; this suggests that Windows NT Server’s single environment for applications and network services simplifies software installation and enables IT to leverage its skills.

Outside services for training and installation/systems integration complement internal labor for some companies; others may outsource all installation/integration activities. Since use of these services is not universal, Cahners In-Stat Group weighted these out-of-pocket costs based on the rate of use of these services within each environment for each size class. Overall, use of outside training services is more widespread among Windows NT Server users, while NetWare / UNIX users are more likely to avail themselves of outside integration services.

The average spending for integration services was 57% less for Windows NT Server users. The average spending on training was 42% less. The Windows NT Server cost advantage for integration services derives primarily from a lower percentage of respondents requiring outside services for integration; the price paid for those services by both groups was similar. The Windows NT Server cost advantage for training services derives primarily from the amount spent; more Windows NT Server respondents were using outside training but paying less, on average, than NetWare / UNIX users.

In considering the cost of investing in a new system, it is insufficient to examine just the costs of the system itself. Most companies have significant investments in systems, networks, and even ways of doing business that may not be easily changed. Since the operating system is a fundamental building block of the total distributed computing environment, the impact on the overall computing infrastructure can be significant.

Cahners In-Stat Group includes the following infrastructure upgrades in its analysis of the total cost of ownership of a server operating system:

• upgrades to cabling or wiring,
  
• changes to the network protocol,
  
• hardware upgrades,
  
• server software upgrades,
  
• client software upgrades,
  
• changes to the directory environment, and
  
• business process or workflow reenginering.

The cost-of-ownership analysis includes both out-of-pocket costs and labor costs. Since not all companies require these types of infrastructure upgrades, the costs associated with the changes are weighted by the percentage of respondents performing the upgrades. These percentages range from 50% to 60% of respondents upgrading hardware, cabling, server software, and client software, to 20% to 40% of respondents changing network protocols, directories, or business processes.

On average, Windows NT Server users reported 25% higher spending on upgrades, including both labor costs and out-of-pocket costs than NetWare / UNIX users. The primary sources of the spending difference were hardware and software upgrades. It is not unexpected that Windows NT Server users would report higher upgrade costs, since Windows NT Server users are displacing existing environments more frequently than NetWare users. An analysis of upgrades by size class can not be considered highly reliable because of the limited number of respondents reporting upgrades and the high variability in the cost of upgrading a networked environment. Organizations assessing whether to move to a new operating environment should examine their upgrade requirements carefully to determine the impact on their costs and deployment schedule.

The initial costs:

• initial systems costs,
  
• installation costs,
  
• outside services costs, and
  
• upgrade costs

combine to form the total initial investment required to support each environment. On average, Windows NT Server costs 9% less to get up and running than NetWare / UNIX environments.

The cost advantage is greatest in organizations with less than 50 users (55%) and for the 51 to 499 user size class (28%). There is no cost differential for respondents with 500 to 999 users, and the 1,000 to 3,000 user size class shows a 23% cost advantage for NetWare / UNIX environments when upgrade costs are included. However, the upgrade costs distort the analysis somewhat by size class due to the relatively small number of respondents upgrading their systems; without upgrades, Windows NT Server shows a 10% cost disadvantage in the 500 to 999 user size class and a 14% cost advantage in the 1,000 to 3,000 user size class.

In order to combine the initial outlays with the ongoing spending into the total cost-of-ownership model, the total initial costs are divided by the average time frame companies use to analyze return on investment for server operating systems (two years) to arrive at an annualized initial cost.

The ongoing costs associated with a distributed environment include both basic server management costs as well as the costs associated with running applications in that environment.

The labor associated with the ongoing management of the networked server environment is one of the key drivers of long-term operating costs. While this cost-of-ownership analysis only examines the direct labor costs associated with managing the server environment, the ongoing labor time can also represent an opportunity cost to the business; time spent on the tactical aspects of basic server management can keep an IS staff too busy to be proactive in addressing new business requirements.

The categories of ongoing labor considered in this analysis include:

• configuration management;
  
• backup;
  
• user adds, moves, and changes;
  
• network resource adds, moves, and changes;
  
• LAN/WAN troubleshooting;
  
• security management;
  
• directory management;
  
• client software distribution;
  
• help desk support; and
  
• other LAN/WAN management.

In total, Windows NT Server users spend 27% less time on server management tasks and, therefore, incur less cost than NetWare / UNIX users.

With the exception of help desk support, Windows NT Server sites report a cost advantage over NetWare / UNIX sites for ongoing management tasks. Windows NT Server’s cost advantages are dramatic for organizations with less than 500 users. There is a modest cost advantage (3%) for NetWare / UNIX in the 500 to 999 user size class, and a cost advantage for Windows NT Server in the 1,000 to 3,000 size class (22%).

Applications-related costs represent an increasingly important aspect of server operating systems costs since companies are increasingly relying on servers to run mission-critical applications. The ongoing applications-related costs included in this analysis are:

• ongoing development of new applications,
  
• ongoing deployment of packaged applications,
  
• maintenance of in-house applications,
  
• maintenance of packaged applications,
  
• user access management,
  
• database management, and
  
• user support.

On average, Windows NT Server users report spending 17% less time and, therefore, incur lower labor costs than NetWare / UNIX users for applications-related labor.

For ongoing applications-related tasks, the greatest cost advantage for Windows NT Server lies in the 500 to 999 user size class. Windows NT Server has a cost advantage of 13% to 24% in the less than 500 user size classes, and no cost advantage or disadvantage in the 1,000 to 3,000 user size class. The greatest contributors to the cost differential are database management and maintenance of packaged applications; Windows NT Server had a cost advantage for all tasks.

The total ongoing costs ¾ server management and applications-related costs ¾ are gathered on the basis of the total number of hours per week spent on each task within the respondents organization by all personnel. The total ongoing costs are annualized by multiplying by 52 weeks to arrive at a total annual ongoing cost. The total annual ongoing costs are overall 22% less for Windows NT Server users than for NetWare / UNIX users.

The cost advantage for Windows NT Server is greatest for the 1 to 25 user size class (61%), followed by the 26 to 50 user size class (39%) and the 500 to 999 user size class (24%). The remaining groups report only an 8% cost advantage for Windows NT Server.

The total cost of ownership combines the initial costs with the ongoing costs. On average, Microsoft Windows NT Server environments cost $326K to run each year, while NetWare / UNIX environments cost $389K to run. In order to more effectively compare these cost figures, Cahners In-Stat Group divided the total cost by the average number of servers, sites, and users.

Windows NT Server’s cost advantage over NetWare / UNIX environments is strongest in small organizations, is moderate for mid-size organizations, and is minimal for large organizations. The cost differential is most dramatic on a total basis or a per-server basis; small organizations experience cost advantages as high as 65%. Mid-size organizations see moderate cost advantages in the 10% to 15% range. On a per-site basis, Windows NT Server does offer a significant cost advantage even for large organizations (21%), although the cost advantage is insignificant on a per-server or per-user basis.

As organizations increase in size, the overall complexity of the computing environment impacts the ease of use and implementation of Windows NT Server. However, even the moderate cost advantage of Windows NT Server in mid-size and large organizations is an important finding considering the time Windows NT Server has been on the market relative to NetWare and UNIX products; since businesses have had less time to work with the product, they have had less time to become proficient with it and efficient at managing the environment. Companies are also more likely to require infrastructure changes to incorporate Windows NT Server into their environments. These factors would suggest that Windows NT Server’s cost of ownership should be higher, rather than lower, than NetWare / UNIX environments if the environments delivered comparable ease of use, installation, and administration.

Microsoft’s lower cost of ownership translates directly into greater satisfaction. Seventy percent of Windows NT Server users indicate that they are very satisfied with their server environment; fifty-seven percent of NetWare / UNIX users are very satisfied.

Windows NT Server users are also more likely to indicate that their environment is very easy to implement than NetWare / UNIX users.

For Windows NT Server users, ease of use and implementation is among the top three decision criteria in choosing a server operating system. For both Windows NT Server and NetWare / UNIX users, price/performance and compatibility with existing systems are the top two criteria.

As companies increasingly rely on distributed computing environments for critical applications, it is more important than ever that server operating systems provide a robust environment that is simple to implement and manage. Using a single operating environment for applications and network services enables companies to more effectively leverage the skills of their IS organization and focus more of their efforts on building value and less on basic maintenance tasks. This study clearly demonstrates that Windows NT Server provides these types of advantages over combined NetWare / UNIX environments.