3. Background
Captaris WebMail discontinued
Follow-on product more expensive,
unnecessary features
Primary usage:
Traveling faculty/staff
Home users
Students in campus buildings
Requirements:
IMAP mailbox access
Comparable features
4. Selection Process
Selection team identified alternatives
Hands-on evaluation where possible
Other schools
Products in use
Evaluation reports
Selected IMP from the Horde Project
http://www.horde.org/
5. Horde IMP
Web application suite
E-mail, contact manager, calendar,
portal, task list, notepad, etc.
University of Pennsylvania report
http://www.upenn.edu/computing/eval/2002/webmail/
Other schools using IMP:
MIT
Purdue University
Tulane University
University of Michigan
6. Implementation Plan
Mix of open and closed source
Use closed source mainly to leverage
existing infrastructure
Minimal changes to IMP source
Use extended testing/pilot phases
Gradual deployment
Batch migrate personal address books
from Captaris WebMail to IMP
7. Open Source Components
Red Hat Enterprise Linux 3 AS
Apache HTTP Server 1.3.x
PHP 4.x
Horde IMP, Turba
up-imapproxy
Sendmail
8. Closed Source Components
IBM LDAP Directory Server
Oracle Database Server
User preferences
Personal address books
AIX IMAP Servers
Microsoft Active Directory
IBM Network Dispatcher
Load balancing, fail-over
9. Local IMP Modifications
LDAP directory integration
Default display name
Default reply-to address
Login screen design
After initial deployment:
Quota display
Outlook address book export
10. Oracle Database
Server LDAP Server
http(s)
oci8
`
POP/IMAP
Servers
(Blue)
imap
Web Servers
smtp
User
Outbound SMTP
Servers
krb5
Active Directory
Servers
Network File
Servers
nfs
Network
Dispatcher
Servers
ldap
Open Source
Closed Source
11. Hardware
3 IBM xSeries 335 servers
Early hardware problems
Minimized hardware vendors (IBM
RS/6000 and pSeries AIX servers)
Little value from vendor’s cross-
platform hardware management tools
Greater value from existing
relationship and knowledge
12. Administration
Highly-skilled team
Proactively monitor critical services
using open source, vendor-neutral
tools (Spong, Cricket)
Centralized, vendor-neutral
configuration management (depot)
Immediately test and apply security
patches
Review and apply other patches as
needed
13. Implementation Challenges
Mixing open and closed source (e.g.,
PHP and Oracle)
Sizing/scaling, impact on other
systems
up-imapproxy in early development
Captaris refused to release WebMail
address book structure
Early hardware problems
Difficulty proving hardware vs. software
problem
14. Evaluation — Current State
3M requests/day
24K active users
No further hardware issues
Some performance issues (mainly
impact on other systems, esp. NFS)
15. Success Factors
Management support
Preview installation
Formal project management
Hardware vendor relationship
Vendor-neutral management
infrastructure already in place
Minimal changes to source
Gradual deployment
In-house technical expertise
16. The Bottom Line
Support costs:
External costs
(licensing,
contracts, etc.)
Internal costs (staff
time, lost
productivity)
Commercial
software promises
to lower internal
costs in exchange
for higher external
costs Closed Source
External
Costs
Internal
Costs
17. The Bottom Line
Reality:
Commercial
software doesn’t
always deliver
Switch to open
source:
Internal costs may
decrease, or
increase only
slightly
External costs may
decrease
significantly Closed
Source
Open
Source
External
Costs
Internal
Costs
18. Benefits
Decreased operating costs (licensing,
software support)
Able to meet unique integration
requirements
Easy to add, modify features
Display Blue quotas
Outlook address book export
19. Ongoing Challenges
Mixing open and closed source
Simplest when based on common
libraries and standard protocols (e.g.,
imap and ldap)
More difficult when using closed source
libraries and proprietary protocols (e.g.,
Oracle)
Sizing/scaling
Planning for performance
Predicting impact of changes to current
environment
20. Conclusions
Significant savings possible by cutting
external licensing/support costs
Mitigate effects on internal costs:
Local technical support expertise
Use proven technology (applications,
libraries, protocols)
Minimize changes to source
Allow learning and discovery time
Anticipate potential risks