Managing Enterprise Networks: How do we reduce IT costs?

Download Report

Transcript Managing Enterprise Networks: How do we reduce IT costs?

Managing Enterprise Networks:
How do we reduce IT costs?
June 2, 2005
Self-Managing Networks Summit
1
Participants
•
•
•
•
•
•
•
•
Hari Balakrishnan
Rebecca Isaacs
Lee Gates
Venkat Padmanabhan
Sunjeev Pandey
Jennifer Rexford
Dave Thaler
Alec Wolman
Self-Managing Networks Summit
2
What IS an enterprise network
anyway?
• Network is a cost center, business is something
else
• Same company owns infrastructure and end
systems
• Security is important
• Apps are often more constrained
• Heterogeneity is common
– Data center? Extranet? VPN? External face?
– Many different types of devices and systems
• There is no such thing as a “typical enterprise
network”!
Self-Managing Networks Summit
3
Top sources of cost (1/2)
• Configuration
– Software deployment/installation
– Hardware deployment/installation
– Testing and certification of new hardware and software
• Troubleshooting
–
–
–
–
–
–
Intermittent failures
Serialization and disjoint sources of troubleshooting data
Lack of logging standards
Noise in data (too much)
Help desk, telephone support
Attacks and viruses
Self-Managing Networks Summit
4
Top sources of cost (2/2)
• In-house apps/tools
–
–
–
–
–
Alpha code for in-house business apps
Developing in-house management apps
Lack of sufficient tools motivates in-house apps/tools
Own unique set of problems
Web site and wiki maintenance
• Training
• Heterogeneous versions across devices/hosts
• Extranets, remote access
Self-Managing Networks Summit
5
Ways to reduce cost
•
•
•
•
•
•
•
•
•
Increase resiliency - redundancy and self-adaptation
Homogenize
Better ways to test and sandbox
Better models to understand complexities and
dependencies
Data consolidation and mining in real-time
Proactive monitoring before help desk calls
More self-help tools/systems
Sharing of problem/workarounds - knowledge base
Have a backup system 1 version behind
Self-Managing Networks Summit
6
Barriers to adoption
• Pace of change - new technologies, apps,
hard for automation to keep up
• Value proposition for investment
– Can savings be quantified?
– Short-term (switching cost) vs long-term cost
tradeoffs & budgets (CapEx vs OpEx
approval)
• Awareness of tools & best practices, job
security
Self-Managing Networks Summit
7
Next steps for research
• Study of enterprise attributes for various types of
enterprises
• Highly reliable systems out of homogeneous
components at low cost
• Standard evolvable logging schema to help correlation
and aggregation problems
• What is a “normal” model based on captured data
• Performance debugging
• Network aware apps, what is good behavior
• What-if modeling and sandboxing of complex systems
• Methodology for interactions between protocols
• Scalable network emulation
Self-Managing Networks Summit
8