Common Solutions Group: StoneSoup.org
Common Solutions Group: StoneSoup.org
This paper summarizes and expands upon the Common Solutions Group Network Workshop, Network Financing discussion, October 11, 1995, (participants listed at the conclusion of the paper) and the CSG Network Survey conducted in the spring of 1995.
Special thanks to David Wasley, UC Berkeley, for his assistance, particularly with Section VII. Financing Principles.
electronics become obsolete quickly; typical useful life is 3-5 years
conduits, cabling, wiring closets and similar investments should last the life of a building; typical useful life is 10-25 years
Attendees agreed that campus digital network usage has been doubling approximately every year for the last 10 years. CSG institutions expect this rate of growth to continue for the foreseeable future, if not increase (see notes on possible discontinuities later in this paper).
Both the increase in scale of digital networks, the increase in diversity of services (e.g., 10baseT, 100baseT, ATM to the desktop) and the rapid obsolescence of equipment, digital networks are growing increasingly complex producing corresponding pressures on staff and support costs.
During a growth phase, financing must be adequate to cover substantial investment in new backbone and station wiring; equipment, design and engineering; end-user installation; and initial support efforts;
Once a network stabilizes, financing must cover routine equipment maintenance, equipment replacement, capacity growth, and so on; and
Neither patterns of network use and nor underlying network technology are yet mature, so even a stabilized network must plan for possible "discontinuities" due to dramatic changes in demand or technology (see further discussion below).
End-user incentives must line up with institutional incentives (e.g., don't make it expensive for end-users to choose the technology most appropriate or cost-effective for the institution); and
End-user incentives can materially aid or inhibit institutional strategies (e.g., charges requiring significant end-user trade-offs to afford will greatly inhibit achieving ubiquitous connectivity).
- Core and station electronics acquisition, maintenance and replacement
- Core network design and engineering staff
- Core cabling maintenance
- Station electronics, wall jack activation (includes hub but not desktop electronics)
- Inter-building cabling costs
- Intra-building station cabling and maintenance, wall jack installation and wiring closet construction
- End-user support
- End-user software baseline
If not included in the expense base, these expenses are generally paid with one-time money (e.g., intra-building cabling) or left to end-users to provide (e.g., end-user software and support).
Recognizing many variations in the expense base, a thumbnail survey of CSG institutions at the networking workshop established the unit cost of a combined voice/data connection at roughly $40/mo. +/- 20% (typically $15-25/month for existing voice charges, $25-15/mo for wired or dial-up data connections). This result was obtained despite vastly different assumptions across the institutions and thus cold represent more of a price point that clients are willing to pay than an accurate estimate of costs.
The typical baseline of network services, where applicable, included:
- connection, IP stack
- universal email
- network ID
- administrative client/server application clients
- file server access; technology for printing
- web browser; internet suite
Some schools include the desktop computing device in or as an adjunct to the network financing model.
Models for network financing cover a range of options from viewing the network as a strategic resource to viewing the network as a mature business.
The strategic resource model holds that the network is an essential underlying element for critical institutional goals and thus must be deployed in advance of demonstrated demand to allow the early development and implementation of institutional applications dependent on full network connectivity, capacity and reliability. The prototypical example of strategic network financing is NSF's funding for NSFNET from 1988 into 1995.
Central funding is established at a point in time. During the rapid growth phase of a digital network, the expenses of the network will grow much faster than normal institutional budgets, so a funding shortage can develop unless the financing model has allowed for this growth.
The strategic model generally establishes network connectivity in advance of some end-users' ability or interest to use it. The full value of this investment is realized only if institutional applications that take advantage of universal connectivity are actually deployed in conjunction with the network installation.
UC Berkeley funds core network infrastructure centrally and during the high growth phase of the network also substantially subsidized end-user connections. To meet the expenses of rapid growth which central funding has not kept pace with, Berkeley recently moderated the end-user subsidy.
Stanford funds core network infrastructure centrally and requires departments to organize and fund end-user connections (central help for departments is available at cost).
The business model holds that the network is basically a mature technology and thus must be deployed with costs directly proportional to benefits and funded by full cost allocation to the beneficiaries. The prototypical example of the business model is many universities' departmental charging for telephones.
Only those end-users who perceive that networking is worth the cost connect, inhibiting the implementation of institutional applications which require comprehensive campus connectivity; and
Rate and charge-back issues present a significant administrative effort.
MIT has essentially achieved full campus connectivity. While central and station cabling were paid for centrally as part of a phone-switch installation, all current operating costs for campus network services are billed to end-users at rates set to fully recover costs.
Yale is still in a network growth phase, but projected expenses at full campus connectivity and has set rates to what they would need to be at full connectivity to achieve full cost recovery. With minor exceptions, the expense base includes all operating and intra-building cabling expenses and the cost of preparing spare telephone station wiring for data use.
Options for financing include central budget allocation and end-user charge-back. Most schools use a combination approach.
Funding strategies of the strategic resource model vary from full central funding of the network, to funding of the "core" network infrastructure that is in common across all users with end-users paying some or all of their "marginal" costs for connecting to the network.
Whatever approach is used, schools increasingly recognize that the telephone, data and video networks share a common cabling plant and are beginning to pro-rate the expenses of phone, data and video cabling. Some schools (e.g., Brown, Cornell, Virginia) are beginning to install voice, data and/or video in a common wall-plate/installation. Essentially all CSG schools believe that these networks will share at least a common cabling infrastructure in 5 to 10 years - the decision as to how rapidly to integrate current common functions varies school by school, but schools are clearly headed toward a single charge for voice/data and video connections.
Schools that apply some charge to digital networks vary on what they base their charging on.
Usage charging is analogous to long distance telephone charging: charges are based on the data transmitted or some other measure of usage. Essentially no school charges by usage.
Capacity charging is analogous to local telephone charging: charges are based on a unit of capacity, whether or not it is used. Typically capacity charging is based on connections, either connections of an individual desktop machine, a LAN or a departmental network. Measures of connections include: active wall jacks, assigned IP numbers, router ports, etc.
Some schools (e.g., MIT, Princeton), charge based on end-user connections (e.g., IP address or node), but provide a cap on charges by an administrative unit (e.g., department).
Not everyone includes end-user installation and support. Some schools link a strategic funding model with providing desktop computers for faculty and staff.
Installation costs arise from: installation or preparation of station cabling and wall jack, station electronics and jack activation.
Installation charges can pose a substantial obstacle to the growth of network connections. Institutions which see networking as strategic generally either subsidize installation costs directly, or build them into the ongoing charges for the network.
The Network Workshop revealed that essentially all CSG schools have a technical direction adequate to meet current levels of growth in network usage within current financing for the foreseeable future (two years). However, all schools are worried that a major change in demand (e.g., a new "killer application" or expectations (e.g., a telephone system level or reliability) have the potential to greatly expand expenses and therefore break current financing models. The following are the primary issues of concern:
A dramatic change in the definition of the network, for example, a decision to move from medium to high reliability, can significantly increase network costs (introduces the need for redundancy in cabling, electronics and core staff).
While many schools identify a "telephone system" level of reliability as a goal for the digital network, the technologies making up the digital network are not nearly as mature as telephone technologies. The digital network itself is vastly more complex than a telephone network and may include the direct connection of end-user devices which share or influence network routing and other reliability affecting services. John Leong (CMU) argues that the complexity of attached equipment and the constant change of data networking equipment, protocols and configuration means the data network simply cannot be as reliable as today's telephone networks; this level of reliability will become possible only when the network technology itself matures and the pace of change of the network infrastructure moderates.
The broad adoption of World Wide Web technology represents one application that dramatically increased the capacity requirements for campus networks. Sometimes such applications are institutional, e.g., digital libraries, in which case the cost of the needed new capacity can be considered as part of the application (though it usually is not). This application based cost/benefit analysis for investment in the network contrasts with the usual imprecise basis for forecasting network demand which, if insufficient, results in rationing through limited network performance.
The emergence of a new network based application which requires substantial bandwidth to support (e.g., real time video conferencing or voice over data) will overwhelm current network capacity.
It is essential that new network based applications consider the implications for network capacity when they are conceived and implemented.
Over the last six years, digital networks have progressed more or less gracefully from shared ethernet with routed backbones to switched ethernet at the desktop to collapsed backbones or switched ethernet or switched fast ethernet at the backbone. Based on current growth rates, a new backbone technology will be needed in 2-5 years time.
An example of demand driving a need for new technology would be the need by some end-users for high-quality digital service at the desktop, requiring the introduction of new technology at the desktop such as ATM.
Such needs for new technology could introduce significant transition costs which may not be adequately met by current annual equipment replacement budgets.
As the network grows more complex, it is possible to create an environment that becomes unmanageable; this is particularly true when the complexity grows along multiple dimensions, e.g., the physical complexity of the network and multiple protocols. Consensus is the best approach to this problem is to avoid it, and, to this end, several schools have elected to restrict their campus networks to TCP/IP protocols only .
Institutional Strategy
The network infrastructure is a critical part of any institutional information technology strategy. This strategy must be developed and the network financing structure then constructed to promote that strategy. Institutions which develop network financing ad-hoc will almost always fail to achieve strategic information technology goals.
A typical strategic goal would be to rapidly and fully deploy the network for use by all faculty, staff and students; the network financing model will have substantial influence on an institution's ability to achieve such a goal. The following sections provide specific examples of this principle.
Scaling
Networks and their expense bases are growing rapidly, so any successful financing structure must scale with that growth. Rapid and successful deployment of a network can only take place if adequate resources are available to support that growth.
End-user and departmental incentives
The financing model must provide end-users and departments with incentives that align with institutional goals. For example, a campus will greatly inhibit the deployment of network connections to all faculty, staff and students if user fees provide a strong disincentive to connect or encourage departmental rather than institutional networking.
Training and support
Creating a physical network is not sufficient. The financing model must support adequate training and support for network use by faculty, staff and students as well as creating the physical network.
Attendees: Eric Aupperle, John Charles, Dick Cogger, Tony Conto, Mike Dorl, Larry Dunn, Doug Gale, Bill Graves, Terry Gray, William Green, Steve Hall, Paul Holbrook, Greg Jackson, Ken Klingenstein, Dave Koehler, John Leong, George Loftus, Phil Long, Mark Luker, Jack McCredie, Bob Morgan, Will Murray, Erikas Napjus, George Pipkin, Sam Plice, Joe Paolillo, Karl Reuss, David Richardson, Mike Roberts, Amanda Rushing, Jeff Schiller, Tim Sigmon, Steve Updegrove, Russ Vaught, Dave Wasley, Don Wolfe, David Wood, Steve Worona.