Overview
The Email Subcommittee was formed by the UCSB Campus
Network Committee and charged with recommending a solution for
undergraduate email. As part of this effort the Email Subcommittee
formed a subcommittee to survey other institutions similar to
UCSB with the purpose of:
- determining if anyone was using an
email solution that the task force was not already aware of, and
- developing a model for the level of funding and support that
was required to provide campus-wide undergraduate email service.
A brief initial survey was sent out, and those institutions that
responded positively were then contacted for a more detailed survey.
The following twelve institutions were surveyed in this manner.
- UC Office of the President
- UC Davis
- UC Irvine
- UC Los Angeles
- UC Santa Cruz
- UC San Diego
- UC San Francisco
- Cal Poly San Luis Obispo
- University of Washington
- University of Michigan
- Carnegie Mellon University
- University of Georgia
The survey was expanded to include the references
for some of the products that the Email Subcommittee was researching
and/or evaluating. The following seven reference sites were contacted.
- HP Openmail: University P
- A Major Oil Company
- San Bernardino Superintendent of Schools
- ESYS Simeon: Duke University
- University of Maryland
- Groupwise: Georgetown University
- Williams Information Services
Five of the institutions surveyed were in the process
of transitioning from one email system to a new email system.
Survey results reflect their new systems. Most of the surveys
were done by telephone interviews, and the rest were done through
email.
Survey Results
A. Components of the Basic Service and System
Nine of the institutions used a centralized system,
five used a distributed system, and four had a system that incorporated
both centralized and distributed attributes.
All of the hardware used were Unix systems from various
vendors including IBM, Sun, DEC, HP, and PCs, except for the two
Groupwise sites, which were using PCs running either Novell Netware
or Unixware. The number of systems used to support these services
ranged from 1 server to 60 servers depending on both the number
of users supported and the server software being used. The number
of users being supported ranged from 1,200 to 55,000. The ratio
of servers to users for each type of service varied as follows:
POP systems ranged from 1:1200 to 1:14000; IMAP systems ranged
from 1:917 to 1:10750; HP Openmail systems ranged from 1:3100
to 1:10100; and Groupwise systems ranged from 1:120 to 1:268.
Some institutions only ran mail on their servers
and did not allow interactive logins; some also ran other Internet
services such as DNS or webservers; and some allowed interactive
logins in addition to running mail and Internet services.
The majority of the institutions support PCs and
Macintoshes, with almost half also supporting Unix workstations.
Three support X-terminals, and four support ascii terminals. Most
sites support more than one client. The clients supported by
our survey sites are: Eudora (7), Pine (7), Simeon (2), other
POP clients (2), Mail Drop (2), Mailstrom (1), Netscape (1), Siren
(1), and Openmail (1). The clients supported by our reference
sites were for the referenced software except that the Simeon
reference sites also supported other IMAP or POP clients, and
one of the HP Openmail references used CC:Mail instead of the
Openmail client.
All of the institutions supported some sort of remote
access protocol, with the majority supporting PPP. One site is
outsourcing their dial-in service to MCI this coming fall, and
three others are looking into outsourcing this service in the
future.
All but one of the institutions conducts official
business via their email system.
Most institutions use some authentication method
in addition to standard unix passwords: five use Kerberos, two
use TACACS, and one each uses a home-grown database, shadow passwords,
RADIUS, and identd logging. Two sites said that they try to have
strong acceptable use policies to deal with problems when they
occur.
The number of users either currently supported or
planned to be supported at the various institution ranged from
1,200 to 55,000, with the average number of accounts being 20,838.
Several institutions said that their system was scalable, and
that they would just add more servers as they needed. One site
said that the number of accounts is limited by disk space. Another
site that provided login access said that the number of accounts
was limited by physical memory.
All but four of the Universities were already supporting
or planned to support faculty, staff, and students. The exceptions
are one that only supports students, one that also supports alumni,
and two sites where students were on a completely different system.
Ten of the institutions impose disk space limitations
on their users, and the quotas range from 1 to 15MB. One also
limits CPU and connect time. Five of the sites don't want to impose
disk space limits, and two of the HP openmail references would
like to, but the current version doesn't provide that functionality.
The transitioning sites felt their old systems didn't
scale, but they planned on the new system scaling. All of the
other sites felt that their systems scaled well. Several sites
had encountered I/O bandwidth problems, two had problems updating
huge account databases. Other problems mentioned were: NFS &
mirroring; handling the load of mass mailings and large attachments;
no redundancy for disk failures; needing better administration
tools; needing better clients; and underestimating the load required
by some of the clients.
All of the institutions offer a variety of Internet
services including news, web browsing, ftp, and telnet. Eight
of the universities are also providing student web pages, but
two of them are not allowing CGI scripting.
B. Access Issues
The numbers reported for concurrent user access depended
on whether the site allowed logins to the server or only POP or
IMAP access. The login servers had between 100 and 300 concurrent
users. No numbers for concurrent users were given for the POP
servers. Three of the IMAP servers were reported with 200 concurrent
users, and one site was hoping to get 500 from their IMAP server.
An HP Openmail server was reported to have 400 concurrent users.
A survey done by University of Georgia reported 120-150 concurrent
users for must Unix systems, and 250 for multiprocessor Suns.
All but one of the institutions provides access to
their system from public access labs, desktop systems, and home
systems. The exception only has access from desktop systems.
The users either use telnet, POP or IMAP to connect to the server
depending on the service being provided. For two of the POP institutions,
the students need to carry around floppy disks which they use
to receive downloaded mail at the public access systems. Six
of the universities have wired at least some if not all of their
dorm rooms.
The total number of dial-in modems ranges from 8
to 1300. The ratio of user accounts to dial-in modems varies
from 21:1 to 750:1. The total number of public access workstations/terminals
ranges from 80 to 1500. The ratio of user accounts to public
access seats varies from 6:1 to 275:1.
C. Facilities and Support
Three of the institutions estimated their costs for
hardware to be less than $100,000, they could do this by taking
advantage of hardware hand-me-downs or special vendor trade-in
programs. Three institutitions said their hardware costs were
between $100,000 and $200,000, and four institutions estimated
that their hardware costs were between $200,000 and $300,000.
Two of the institutions didn't really have a budget,
but they were able to get what they needed because it was a critical
campus resource. The only budget numbers estimated were $70,000
per year and $200,000 per year. Most of the people surveyed didn't
know what the budget was for supporting electronic mail.
Ten of the sites provide classes in addition to documentation
either in hardcopy or on-line form. Four sites just provide documentation.
Three also use their help desk.
Thirteen of the universities provide their email
service for free to their students. Two of the universities charge
some fee, and two more are considering charging a fee. One of
the universities used to charge a fee but they abolished it.
Email was supported by staff dedicated to that function
at some institutions, and it was just one of several computer
services supported by the staff at other institutions. The staffing
numbers should also be considered along with the number of accounts
and systems being supported as well as the email technology being
used. The staffing numbers are reported in this context in a
chart that follows this report. There is also an appendix that
provides more detail to how the staff is allocated at each institution.
Conclusion
None of the solutions were perfect or entirely trouble
free. But each site was happy with their solution, although they
also each had things about it that they wanted to change or add.
The sites using an IMAP server strongly believe that
IMAP is the required technology for providing a campus wide email
service. Two of the sites surveyed have developed their own IMAP
server. They also desire more mature clients for the PC and Macintosh.
The POP sites were very happy with their systems and felt that
the Eudora client had superior capabilities to other email clients.
But they are starting to have issues when faculty and staff want
to access email from an office system and a home system. The
students are considered 100% mobile and download their mail to
a floppy disk that they take with them. The HP Openmail references
seemed to be the most enthusiastic about their system, but they
were all waiting for the quota capabilities available in the new
release. The Simeon references were positive about the IMAP architecture,
and felt that although the Simeon client was young, it keeps improving
with each release. The Groupwise references also liked their
product. Their environment was extremely distributed with no
more than a few hundred user accounts on each server. They were
also looking forward to the new release that will have a client/server
architecture.
Appendix
The appendix lists the institutions surveyed, and how each allocates personnel for maintenance and support of email systems.
