Academic Computing Coordinating Council: Remote Access Project

AC4 Home | Meetings | 4/19/99 | Remote Access Project

Remote Access Project (RAP)

Summary

Permanent funding and better availability of service are urgently needed for remote access to the Campus Intranet and the Internet through the telephone system. It is assumed that the present "14.4 K" modem pool will be replaced by "56K" modems for the use of staff and students in the very near future. Several proposals for further provision of remote access are presented and several funding models for the ongoing costs are proposed. In general, it is recommended that the faculty modem pool be continued and increased as necessary to maintain high availability. It should continue to be funded centrally. One-time central funds should be used to improve the student modem pool and the ongoing costs should be recovered by an increase in the registration fee of $9 per quarter to $22 per quarter depending on the model. Similarly, for staff, one-time central funds should be used to improve the modem pool. Departments should decide who is given remote access. Ongoing costs should be borne by either departments or central funds or shared between these sources. An agreement with an ISP should be explored for department-sponsored access from beyond the Davis local calling area.

Introduction

Remote access refers to computer connections to the Campus Intranet and the Internet from off-Campus. Students need remote access to complete and submit required coursework, and to contact instructors by email, because on-Campus facilities are inadequate. Faculty members need remote access when they travel on business and when they work outside normal working hours, e.g. answering student emails in the evening (faculty workload study showed workload averages 60 h/wk). Staff members need remote access for approved telecommuting, overtime work outside normal working hours, and when their University duties take them off-Campus.

Remote access was originally needed by only a small subset of the Campus population and it was provided without charge to end-users and subsidized from telephone charges. As the use of remote access grew and telephone service became more competitive, the subsidy became untenable. Remote access still does not have a permanent source of funds, although it has been supported by ad hoc infusions of one-time funds, most recently about $500,000 for 2-years (1998-00).

The most recent comprehensive recommendation was made by the JCCIT (Joint Campus Committee on Information Technology) about 2 years ago. JCCIT recommended that the Campus modem pools be phased out and that users purchase remote access in the commercial market (through an ISP [see Note 1]). This recommendation was adopted but didn't work. Only a few users purchased remote access and they found themselves excluded from significant portions of the Campus intranet (due to domain-protection for copyright and other reasons). The growing demand for remote access and the static or declining number of Campus modems available led to saturation of the modem pool.

Currently, there are four groups of modems on Campus (see Note 2), plus the ISP option:

Student/staff modem pool, 376 modems, contention ratio (customer base/modems) 76, very poor availability (figure 1). This pool is planned to increase to about 815 modems very soon, reducing the contention ratio to 35. The actual users per modem are now 30 and will fall to 14 but these figures will rise as the number of users increases up to the values for customer base per modem.
Faculty modem pool, 96 modems, contention ratio (customer base/modems) 16, excellent availability. The actual users per modem are now 8 but this will rise as the number of users rises up to the customer base per modem value of 16.
Legacy modem pool, about 500 modems, will be phased out very soon (currently only used for short, 20 min, sessions).
Distributed modems, about 700 or so. Availability limited to users designated by departments or units. Threaten the integrity of the voice system (see Note 3); may be a security risk.
ISP: a subset of each group-students, faculty and staff-on Campus has ISP accounts, which generally provide excellent availability at a monthly cost in the $10 to $20 range.

The ISP option is the only way for students to obtain good access. For faculty and staff, it is a cost-effective choice when long-distance telephone charges would be incurred by dialing in to Campus. Faculty members have excellent access through their modem pool. Staff members with distributed modems have excellent access; others have very limited access through the student/staff pool, or good access through an ISP.

Figure 1. Daily use of the student/staff modem pool. Data for April 14/15, 1999. Review additional data at http://noc.ucdavis.edu/mrtg/iras.html.

There are two fundamental problems with the present system.

There is no permanent funding model, making planning impossible.
The existing student/staff modem pool is inadequate and a major hindrance to the efficient functioning of the University.

Now that remote access has become an essential part of doing business in the University, a solution has to be found that is an acceptable compromise between availability of remote access and cost, both to end-users and to central funds.

Process

The Provost charged the Division of Information Technology with developing a sustainable plan for remote access and the Division produced the RAMP report and presented it to the AC4 (Academic Computing Coordinating Council). AC4 consulted with AdC3 (Administrative Computing Coordinating Council) about the staff modems and AdC3 has taken over the responsibility for recommendations in this area. AC4 remains responsible for recommendations for students and faculty. A joint committee to combine the recommendations has been proposed but not yet implemented.

AC4 developed the following set of academic criteria against which to judge proposals for remote access.

Students, faculty and staff need to access the UC Davis Intranet and the Internet from their homes.
Arrangements for remote access should support rather than discourage the use of the Internet for academic purposes.
Faculty members should be free to incorporate reasonable use of the Internet and Intranet in their courses as needed without being constrained unduly by lack of student access.
Research should not be constrained unduly by lack of access to the Internet and Intranet.
Administration should not be constrained unduly by lack of access to the Internet and Intranet.
Remote access policies should be consistent with UC Davis' network security policy.
Remote access policies should encourage cost-effective use of remote access facilities.
All students should have the access that is needed to complete the courses in which they enroll.
A university subsidy, if any, for remote access must be consistent, in nature and amount, with the university's mission.
The amount of funding of a university subsidy, if any, for remote access must be appropriate in the context of the overall needs of the campus.
Internet access is used both for university business and other activities. The university should not subsidize use of the Internet and Intranet for non-university business or pleasure.
Consideration of funding models should include efforts to minimize the total cost to the campus and its constituents.
Provision of access and its funding should remain viable as the use of the Internet and Intranet continues to increase.

Using AC4 criteria to select models. Note: AC4 is the score obtained using the mean weightings. Max is the highest score obtained using the individual weightings and Min is the minimum score using the individual weightings

Figure 2. Using AC4 criteria to select models. Note: "AC4" is the score obtained using the mean weightings. "Max" is the highest score obtained using the individual weightings and "Min" is the minimum score using the individual weightings.

AC4 members were invited to submit a personal measure of relative importance (weight) for each criterion and a Council average weighting was calculated from the weights submitted. All models for remote access were assessed for compliance with the criteria and a normalized weighted average was calculated for each model. These averages were used as an initial screen to select models that matched the criteria.

Figure 2 shows selected results of applying the AC4 criteria including evaluation of all existing UC campuses. "Table 1" represents the model shown in figure 3 and Table 1. "Table 2" represents the model shown in Table 2.

Models

Because of uncertainties in predicting usage and technological developments, it is recommended that whatever plan is adopted be reviewed annually and fine-tuned in the light of experience.

One fundamental issue is the inverse relationship between quality of service and cost. Quality of service relates to the probability of connecting when the user chooses and the speed of the connection once made. The probability of connecting depends on how many users are competing for the modems in the pool. This is measured as the contention ratio, which is defined in this report as "customer base/number of modems available" (see Note 5). For example, if there are 1559 faculty members and 96 modems in the faculty modem pool the contention ratio is 1559/96 = 16 (sometimes written 16:1). All the models considered here use a connection speed corresponding to a "56K" modem and session limits (see Note 6) of 2 or 4 h. but no other limits on use.

Figure 3. Cost as a function of contention ratio (see Note 7).

Figure 3 shows how the cost of separate modem pools for faculty, students and staff varies with the contention ratio.

What contention ratio do we need? Clearly the current contention ratio, for the staff/student modem pool, of 76 is inadequate, even with a 2-hour session limit (see figure 1). On the other hand, the faculty modem pool is working well with a 4-hour session limit. The contention ratio for this pool is currently 16 and the pool is rarely more than 50% saturated. However, only about half the eligible faculty members have signed up successfully for this pool -- so there are only about 8 users per modem -- and usage is likely to increase. Consistent with these data, the IT staff suggest that a contention ratio of 15 would maintain good availability of the pool, especially with a 2-hour session limit. For excellent availability, a contention ratio of 7 would suffice, with a 4-hour session limit.

A simple model would provide the same contention ratio for all users. Figure 3 shows that this would cost, annually, about $1.1 million for a contention ratio of 15 or about $1.9 million for a contention ratio of 7.

Suggested phased improvement in contention ratio

A more-complex model would keep the faculty and student/staff modem pools separate, as they are now. Both pools would be increased to reduce the contention ratio to 7 within 5 years and then stay constant. Figure 4 shows how the contention ratio would change. There would be an additional startup cost to add 681 modems per year, for 5 years. The number of users opting for an ISP will affect the overall costs for this model as would other additions, such as a premium pool, or subtractions such as restricting the number of staff members who have remote access. The costs without taking these factors into account are summarized in table 1.

Another alternative is to have separate pools for students, faculty, and staff with different contention ratios. One version of this option, which has been worked out in detail, has a student pool with a contention ratio of 15, a faculty pool with a contention ratio of 7, and two staff pools with contention ratios of 7 and 15. The costs of this model are summarized in table 2. The lower costs of this model are associated with 100% higher contention ratios for most users.

Funding Sources

One of the basic assumptions of this document is that billing to individuals is very expensive in relation to the cost of the services and would not be cost-effective.

This leaves departmental funds, which could be billed through the telephone billing service, student registration fees and/or technology fee, and unspecified central funds.

For students, the approximate annual cost is $700,000 (contention ratio = 15) to $1.4 million (contention ratio = 7). Per student, this translates to about $13 to $22 respectively, per quarter for 3 quarters per year. If central funds were used for the startup costs, then these annual costs, or a portion of them, would probably have to be recovered from a registration fee increase. If the contention ratio were decreased over 5 years from 35 (which is anticipated later this year) to 7, the registration fee increase would start at about $9 per quarter and rise to $22 per quarter in 5 years.

For faculty, probably every faculty member has a good case for needing remote access based on working hours considerably greater than the normal workweek. In this case, departmental billing would basically amount to a tax on departments over which they have no control. It is recommended that faculty access be paid from central funds. The annual costs are in the range $98,000 to $120,000 depending on the contention ratios for the faculty pool and the other pools.

For staff, the approximate annual cost is $260,000 (contention ratio = 15) to $480,000 (contention ratio = 7) (see Note 8). There is a subset of staff who have a clear need for remote access and another subset with little or no need for remote access. Only departments can readily decide which group a particular staff member belongs in. Several proposals have been made for funding staff access including the following

Charge departments for use of a "good" staff pool (contention ratio = 15), which would cost about $37 per user per year. Provide a "premium" pool with a contention ratio of 7 at a cost of $77 per user per year. This places all the costs (except startup) in the departments, where the decision on who should get access lies. This would promote cost-effective use of remote access. However, it may be a significant burden for some departments and it may encourage departments and units to continue to maintain and expand their distributed modem pools.
Provide central funding for the "good" pool and only charge departments for the "premium" pool. The problem here is determining who should have access. If departments decide, they have no direct financial incentive to restrict access, which will reduce the cost-effectiveness of the use of remote access. On the other hand, the zero cost will provide an incentive for departments to discontinue the maintenance and expansion of distributed modem pools (although they are not likely to disappear quickly because it will take time for people to learn that the central system is reliable and available). Access to the "premium" pool would still be charged to departments. The annual charge to central funds for the "good" pool would be up to $260,000 depending on how many staff members were provided with access.
This is a variation on model 2 in which there would be "cap" on the amount of central funds used to support the "good" staff pool. This model requires some way of rationing the subsidized access among the departments.
This is another variation on model 2 in which there would be a partial subsidy from central funds. For example, a 50% subsidy would reduce the department's annual cost to $19 per user while requiring a central subsidy of up to $130,000. There might be a subsidy level that would be fair to departments without overburdening central resources.

These four models assume that a contention ratio of 15 will meet most staff members' needs. If the model chosen tends to restrict the number of staff users to those with heavy remote access needs, then a smaller contention ratio may be needed. Maybe the models for staff should build in a phased convergence of the "good' pool towards the "premium" pool with the aim of providing only "premium" service in, say, 5 years. This model is illustrated in figure 3 and table 1.

Long-distance remote access

None of the above models takes account of users' telephone charges. Within the local calling area, there is currently no incremental charge. However, those who travel on business or live outside the local calling area will have to pay toll charges. Frequently, this will be avoided by using an ISP with local calling numbers for the user's home or destination. Departments can provide this for staff and/or faculty members who need it for their University business. Because having an ISP will usually remove a user from the Campus modem pool, any central subsidy of the pool could be redirected to reducing the cost of the ISP to departments.

Outsourcing

At this time, ISP rates are high compared with the cost of providing remote access internally. Thus, it is recommended that the Division of Information Technology provide remote access for the local calling area. This issue should be revisited when new technologies, such as DSL, become widely available.

Acknowledgements

This report relies heavily on information supplied by the Division of Information Technology, which provided excellent support for the preparation of this report.

NOTES:

ISP means a commercial Internet Service Provider. An ISP usually provides excellent availability for $10 to $20 per month, although the speed of connection to the UCD Intranet could be reduced by Internet delays. An ISP can be used to avoid long-distance phone charges. Note: some UCD materials are inaccessible from an ISP (because of domain protection). (back)
The figures quoted in this report use today's dollars and populations projected for 20001/02. (back)
The voice switch, which connects the many lines on Campus to a smaller number of "outside" lines, is designed to handle the short call-duration of voice messages. The longer call-duration of modem sessions exceeds the design parameters of the switch and will lead to saturation of the voice system, delaying emergency calls as well as others. (back)
Assumes adding 25 modems/yr to faculty pool and 652 modems/yr to the student/staff pool. (back)
Note that only about half the faculty members use the faculty modem pool, so the current ratio of users to modems is about 8. Customer base is used to calculate contention ratio in these models so that they will remain viable as the number of users grows up to the customer base, which is the maximum number of users in a given category. (back)
A session limit means that once connected, a modem may stay connected for the specified time but will then be disconnected. The user may attempt to reconnect immediately but (s)he will then be back in competition with other users until (s)he is connected again. (back)
Calculated using Doug Hartline's formula and assuming all the pools have the same contention ratio. Mixed contention ratios would change the figures slightly. The formula includes maintenance of the pools, replacement, support, and Internet access charges. Factors for inflation (staff salary increases, potential internet fee escalation) or potential deflation (lowering equipment costs) are not included. Startup costs are not shown. (back)
The actual values will vary somewhat depending on the contention ratio in existence for the other modem pools. (back)