Chapter One

Chapter Two

Chapter Three

Chapter Four

Chapter Five

Chapter Six

Chapter Seven

Appendix

Java Computing in the
Enterprise Whitepaper
(Postscript Compressed
141KB)

Java Computing in the
Enterprise Whitepaper
(Acrobat 124KB)

A quick expense analysis of Java Computing based upon the points discussed earlier above can be conducted for a range of Java device deployment percentages. The chart in Figure 5 shows the typical yearly cost per seat of some of the various kinds of desktops discussed in Section 5.

The most common enterprise desktop environments today are either "custom fat client" (variations in the configuration on each desktop) or "standard fat client" (a common desktop configuration defined and enforced by corporate policy). The yearly cost savings of converting some fraction of these desktops to Java devices is analyzed below.

One additional assumption is necessary for this analysis. The one-time cost of a major upgrade of a fat client (e.g., from Windows 3.1 to WindowsNT) has been studied by many CIOs, who have found a wide range of costs associated with the upgrade. An expense level that seems to be supportable by most is in the range of $5,000 per seat. This includes the cost of the software, associated hardware upgrades required, dedicated support costs and end user time consumed in the upgrade process. At the end of this upgrade process, the CIO has added little new capability to the company and has spent a large portion of the MIS budget to implement such an upgrade.

We suggest that, instead of investing in the next upgrade to another fat-client operating system, the enterprise invest that same money into Java migration and quickly begin to capture the savings that can be gained through the Java enterprise computing architecture.

Figure 6 shows the percentage yearly cost savings that can be achieved by converting desktops to Java devices. Two scenarios are shown -- one in which the fat client cost per seat is $15,000/year and one in which the fat client cost per seat is $10,000/year. The first case would correspond to the "custom fat client" enterprise environment noted above, and the second case corresponds to the "standardized fat client." The percentage savings are shown plotted against the percentage of seats that can be converted to thin-client desktops.

In the optimal case -- 90% of all desktops converted to thin client -- the saving is 75% in an enterprise with custom fat-client seats currently deployed and 67.5% in an enterprise with a standardized fat client environment. Individual enterprises will have different conversion rates, which may start small but grow over time. Even at relatively low rates, the savings are evident.

The primary drivers behind Java Computing savings include:

• Reduced acquisition cost of desktop hardware/software system
• Reduced complexity at the desktop
  • All OS and application software stored only on servers
  • All software upgrades done at the servers
  • System administration all done at the servers
  • All data files reside on servers
  • Less on-site support infrastructure required
  • Less user self-maintenance required
• Desktop failures can be corrected immediately with a system swap
• Lower costs associated with supporting fewer platforms
• Lower application development cost
• Lower application support infrastructure cost
• Reduced complexity/cost of network security

The amount of the potential savings through Java is staggering. Sun calculated a possible cost savings for converting 75% of existing fat-client desktops to thin clients in a typical organization (a larger Fortune 1000 company, the current main market for Java Computing), with 100,000 nodes within the organization that are possible candidates¹ for conversion. The following calculation was based on an estimated total cost per seat per year of $10,000² for "standard fat clients" and $2,500 per seat per year for thin clients. (See section 6-1 for discussion on thin client costs.) Such an enterprise could save $562.5 million annually deploying Java Computing.

Using an even more conservative conversion metric of 25% -- which is significantly below the expectations Sun has been given by customers interested in Java Computing -- the savings is $187.5 million. Even this very low conversion rate could produce projected savings that would be seen as a major victory for the CIO in any enterprise. These cost estimates are based on an extrapolation from Sun's internal experience with our 35,000 node TCP/IP network. Cost figures will necessarily vary based on specific circumstances in a given corporation, but we believe the above analysis to be fairly representative. Therefore, we recommend that any medium or large enterprise undertake its own evaluation of Java Computing before investing heavily in upgrading existing PC/Windows fat-client architectures.