Last update at http://inet.nttam.com : Thu Apr 27 22:35:57 1995

Internet Affects the Corporation: Experiences from Eight Years of Connectivity

Suzanne M. Johnson (johnson@intel.com)

Abstract

Intel's experience with Internet connectivity is described in this document. Internet connectivity has emphasized the importance of mail system interconnectivity within the corporation, as well as signified the need for a corporate-wide network architecture. Users have demanded guidelines for usage of the Internet. Each of these areas will be discussed and recommendations made which other large corporations facing some of the same issues, may find helpful.

Contents

1. Introduction
2. Mail: First Use of the Internet
3. Network Architecture
4. Usage Guidelines and Related Business Policy
 5. Marketing Information Available on the Internet
6. Conclusions

References
 Author Information

1. Introduction

Commercial use of the Internet, and connectivity to the Internet by commercial organizations has grown rapidly. Even companies that have been connected to the Internet for years are undergoing major changes in their usage of, and attitudes toward, the Internet. This paper describes the evolution that occurred within Intel, notes characteristics of the evolution which other commercial organizations are likely to face, and explains the process by which Intel resolved, and is resolving the major issues to connectivity and usage of the Internet.

Over the years, we have seen several distinct phases of connectivity. Our first connection, via CSNET, was intended only to provide engineering research organizations with Internet mail capability. We quickly discovered that other organizations had use for the Internet mail facility, and we became painfully aware of the problems posed by multiple mail systems within the company which did not necessarily transfer mail among themselves or with the Internet .

On the heels of exponential growth in mail usage, access to information on the Internet became necessary for a variety of organizations within the company. Our efforts to make this possible led us to a better understanding of the need for an internal network infrastructure that would allow access to information from the desktop, and delivery of information to any desktop within the company. This understanding contributed in a major way to a corporate wide effort to build consensus on a network architecture and network management architecture for the company. These architectures featured technologies that were developed in the Internet environment and grew into corporate respectability. Our network today depends heavily on many of the vendor neutral, open standards such as TCP/IP, DNS and NTP that matured on the Internet. Likewise, hardware technologies such as routers play a similarly important role.

During our many years of Internet connectivity, we have seen that our Internet users want guidelines on how they should use the capabilities available to them. They very much demanded a set of Internet usage guidelines and that those guidelines be available in an easy to access form. This has led to an ongoing effort to define and update guidelines with material appropriate to escalating Internet usage. Having these guidelines in place and easily accessible has become critical as Internet usage became widespread.

In the last year or so, we have entered a new phase of our Internet connectivity. Intel's marketing organizations have decided to have a formal presence on the Internet: an Intel WWW server (http://www.intel.com). This has created a new set of considerations for planning and expanding the Corporation's Internet connectivity. It has also accentuated the importance of marketing's perspective and role in determining the content of such a server.

Perhaps the best indicator of the growth in respectability of Internet usage is seen in the actions of a certain senior technology manager. When we first connected to the Internet, this manager commented that he couldn't understand why we needed Internet connectivity, and wondered if it wasn't just another way for his engineers to get dates and play a wider range of computer games. We recently noted that this technology manager is now a user of a WWW browser, and that he no longer questions the utility of Internet connectivity.

2. Mail: First Use of the Internet

Intel connected to the Internet in 1986 via a CSNET dial-up PhoneNet connection. The connectivity was justified and paid for as a research/development requirement for engineering design and software development groups. At that time, a prerequisite for Internet connectivity was proof that the organization was a bona fide research and development organization. Commercial organizations were expected to refrain from commercial activities on the Internet. Internally, Intel did not offer classes or documentation on how to use the Internet, or even information on the fact that we were connected. Still, we saw exponential growth in the use of Internet mail.

By the end of 1988, our telephone charges for connectivity to the CSNET mail host were approximating the cost of a leased line. By 1989, we were allocating Internet charges among three different Intel divisions: Process Technology groups had joined Microprocessor and Systems development groups as major users of the mail facility.

During this period, Intel's internal network was evolving as four distinct networks [1], better connected in some areas, unconnected in others. There were four distinct mail systems, including one that could not send or receive mail from the Internet. Users demanding mail access to the Internet accelerated the deployment of mail gateways between major mail systems. In fact, demand for a gateway between one of our internal mail systems and the Internet turned out to be so large that we wrote our own gateway, anticipating the vendor's decision to subsequently offer such a gateway as a product.

2.1 Growth in Use of Internet Mail

We began measuring number of mail messages to and from the Internet in early 1987 when mail traffic was averaging a couple of hundred per week. By the beginning of 1988, we were handling about 1200 per week. By the beginning of 1989 it was close to 2500 messages per week, 5000 in early 1990, 10,000 in early 1991. By this time, we had moved off our dial-up PhoneNet connection to a leased line, and mail continued its exponential growth. At the beginning of 1995, we were commonly seeing 150,000 messages per week . Clearly the growth is exponential, and shows no sign of letting up. Groups using Internet mail now extend beyond the original engineering and development groups. Everyone within the company who has an electronic mail address can now send and receive Internet mail. Figure 1 shows the growth in volume of Internet mail as number of messages per week through our primary gateway in Santa Clara.

Figure 1: Growth in Volume of Internet Mail Traffic

It is worth noting that just about every function we have measured that relates to the Internet gateway and Internet usage shows an exponential growth curve when plotted. In fact, this has been so consistent that we have been accused of plotting one exponential curve and then using the same curve with changed labels on the axes of the graph to suit the particular metric under discussion.

2.2 Interest in Information

At the end of 1991, Intel's Library undertook a project to understand how senior engineers, managers and people in marketing create, use, and disseminate technical and business information. Intel Library personnel worked with consultants to identify a sample of Intel knowledge workers with whom to conduct in depth face to face interviews. Twenty-six interviews were conducted at Intel's major domestic sites.

Of the knowledge workers chosen for interviews, general characteristics were that those interviewed tended to be longer service employees (average time 9 years); most were male and most had come to Intel from school or another, non-semiconductor company. Other characteristics of the sample group: most wrote documents for internal rather than external audiences; almost all belonged to one or more professional or scientific organization; most (89%) regularly passed on library information to others and fewer (62%) regularly received information from others.

When the interviewees were asked about their sources for information, 23% identified USENET news as an important source. This placed USENET news on a level with literature searches and slightly above textbooks (19%), as a source for information. When asked what Internet services they used, 73.1% responded e-mail, 23.1% FTP, 11.5% said public domain software, 23.1% indicated USENET news, and 26.9% responded that they did not use Internet services.

Table 1 summarizes responses to a question which asked what services the interviewees would like the Intel Library to offer.

Table 1: Library Services Desired

The interest in information was clearly growing, and the need for an internal network infrastructure that would permit access to information inside the company, as well as outside was becoming very apparent.

3. Network Architecture

Just as the demand for Internet mail arose in each of Intel's internal mail environments, the interest in accessing Internet information sources began to be identified in each of the internal network environments. Although in 1991 Intel supported several network transport protocols (TCP/IP, DECNET, Vines), there was an overall focus on transition to TCP/IP and eventually to OSI. Interviews and discussions during a Library strategic planning effort suggested that key managers and technologists within the IS organizations had developed a good vision of the future Intel network architecture. However, knowledgeable technical staff outside IS did not necessarily share the same vision, and the overall architecture for networking and distributed computing within the company was not understood and supported by all segments of the company.

As one of several information service providers within the company, the Library's strategic direction and choices depended heavily on network bandwidth, network protocols and network management. Since the Library's knowledge worker profile effort had identified access to Internet information as a growing requirement, there were direct implications on the ease at which this could be accomplished in the absence of corporate-wide agreement on a network architecture vision and strategy.

3.1 Identifying Existing Network Related Efforts within the Corporation

Consensus is very important at Intel, and it was necessary to approach the evolution of a corporate network architecture in a manner that would ensure understanding and agreement throughout the company. To this end, we asked consultants to work with us to formulate a scenario planning based activity that would result in understanding and agreement on a single vision for the network. Our first steps involved identifying the extensive work already done in divergent parts of the company on network architecture and planning. We were fortunate in that recent major software development meetings within the company had identified over 200 groups, committees and individuals who were stakeholders in the network architecture. This had been done because there was growing agreement that many other architectural efforts within the company required, or assumed, a highly functional and manageable network infrastructure.

Using this stakeholder information we gathered a significant amount of documentation relating to the existing network and plans for its evolution. We also identified a representative sample of individuals who would need to be interviewed for further information. Our consultants, who were very network knowledgeable, carried out these interviews in person, in so far as possible. This enabled us to collect a good deal of undocumented information and expert knowledge from technologists, users and business managers whose lives and work were affected by the network.

One of the attributes of the consultants with whom we choose to work was their knowledge of computing and communications, and their experience with scenario planning. Technical material gathered, as well as material gathered in interviews was used in building five possible scenarios of Intel's future network. While this was in process, we selected 40 representative members of the Intel network stakeholders' community. We included a sample of key technologists, managers, customers, product developers, and others who felt they had a major stake in the future of the Intel network. We tried to select representatives across divisions and geographical locations. We also identified 50 key network technologists from all areas of the company.

3.2 Network Vision and Strategy Emerges

We invited the 40 key network stakeholders to a two day strategy workshop during which time the consultants led the participants through a scenario planning activity that resulted in consensus on a network strategic vision. The resulting high level vision described a set of standards and services which when implemented would yield a network whose essential characteristic was "rich interoperability". This would mean that enterprise-wide applications, databases, and network services would be accessible from all desktops.

Subsequent to the formulation of the network strategic vision, a network architecture workshop was held with the key network technologists as participants. During this workshop, the consultants identified key decisions that would need to be made to implement the vision, and worked with the technologists to reach agreement on an implementation approach.

A network architecture document emerged from this effort. This document went through two full rounds of review and comment by over 100 people before resulting in a final document.

The result of this effort was an implementation plan for an enterprise internet based on TCP/IP, and many of the technologies and protocols familiar to us from the Internet. The Internet and the internet were beginning to merge.

4. Usage Guidelines and Related Business Policy

From the time of our initial Internet connectivity, users and potential users of the Internet have asked for guidelines for usage. At first we tended to overlook the importance of having guidelines, and having them easily available. However, we quickly realized the necessity of having something available for new users to read. Our initial rationale for creating guidelines, and the guidelines themselves have been described elsewhere [2]. We continue to refine and update these as we learn more about what Internet users are likely to do.

4.1 Posting to Newsgroups

In late 1994, Intel marketing groups discovered the Internet and USENET Newsgroups the hard way. As the online discussion of the Intel Pentium(R) Processor problems unfolded, many marketing engineers learned very quickly about Internet access and how to read USENET Newsgroups.

Later, as the implications of employee posting of messages to USENET Newsgroups became more evident, a group concerned with external communication evolved a corporate business principle relating to such postings. The text of the principle follows:

5. Marketing Information Available on the Internet

Early in 1994, Intel entered a new era in the use of its Internet connectivity. Marketing groups decided that they wanted an electronic presence on the Internet. Work began on a Corporate WWW server, the presence of which was officially announced in February of 1995. A demonstration of the differing perspectives of engineering and marketing groups occurred as implementation of this server proceeded. An engineering development group did the initial work on the technology involved to prepare and update content on the server. Our marketing groups set about converting some of their print program materials for use on the new server. Even though accelerated by working with contractors on the preparation of the material, progress was still not fast enough to satisfy the engineering groups.

In mid-year, the engineering groups decided to turn on the server with content of their own while waiting for marketing to prepare final materials. One of the first pieces of information made available by the engineering effort was a text-based description, written for Intel employees, of comparisons between Intel products and their closest competitors. Marketing felt this was a strange choice for material for this server and redoubled their efforts to prepare information they felt was more suitable.

Marketing groups put a great deal of thought into evolving a unique "look and feel" for Intel's WWW server. This server, plus presence in other electronic communities, has led marketing to rethink the internal processes used to create and publish technical marketing documents. There is a move towards targeting all such documents for authoring in SGML, and then using this as a basis for preparing the various formats required for publication in a variety of media.

6. Conclusions

When we first connected to the Internet, we did not anticipate the impacts that the connectivity would have on the company. We thought we were just providing mail access for a group of engineers and technologists who needed access to universities and industry consortia such as SRC and Sematech.

It quickly became apparent that mail connectivity meant that the mail "must go through" internally and we had to be able to deliver mail to any of the mail systems within the company. This made us very aware of the areas of mail system disconnects.

As mail usage grew, users discovered other useful information sources on the Internet, and demand for access to information from the desktop grew. As we thought through the implications of this demand, we realized that our internal network infrastructure would not easily support this kind of access. This contributed to an effort to establish a corporate network architecture.

Through it all, Internet users and their managers demanded a clear set of guidelines easily accessible and understood. This has lead to ongoing effort to refine and publicize existing guidelines.

Intel may have been fortunate to have connected to the Internet at a time in which it was growing relatively slowly. We had time to work through many of the issues that arose as a result of connectivity. Corporations connecting today will likely find that they are faced with some of the same key issue areas that we encountered. However, resolving the issues may well need to be done in a more compressed time frame. We hope our experience will be helpful to others facing these challenges.

References

[1] J. Sedayao, S. Hambridge, J. Starnes, "Delphi - An Information Resource in a Multivendor Multiprotocol Network Environment," Proceedings of the 1992 ASIS Mid-Year Meeting, Albuquerque, June 1992.

[2] S. Hambridge, and J. Sedayao, "Horses and Barn Doors: Evolution of Corporate Guidelines for Internet Usage," Usenix, LISA VII, pp. 9-16.

Author Information

Suzanne Johnson is manager of the Internet Information Technology group at Intel Corp. in Santa Clara, California, USA. Trained as a chemist and crystallographer, she first came into contact with the Arpanet at Stanford University in 1974 while supporting network collaboration of chemists using the early DENDRAL programs. She joined Intel in 1978, a true believer in the value of network supported collaboration. Convincing Intel to join CSNET in 1985, she has managed Intel's Internet connectivity since that time. She has been a member of the CSNET Executive Committee, the CSNET/BITNET Transition Team, and the CREN Board of Directors. She can be reached at 2200 Mission College Blvd., SC3-15, Santa Clara, CA 95052-8119, USA.

Return to the Table of Contents