Architectural Implications of the

1
Architectural Implications of the FixIt KP
Application

• Jennifer Rexford
• Internet and Networking Systems
• ATT Labs - Research Florham Park, NJ
• http//www.research.att.com/jrex

2
Example FixIt Application

• User problem
• User cannot reach www.example.com
• Asks the KP to fix the problem
• Troubleshooting strategies
• Launch test probes from other vantage points
• Combine problem/success reports across users
• Consider recent anomalies that might be related
• Combine information from multiple domains

3
Probe From Users Vantage Point
10.0.0.1
DNS directs user to different replica of Web site!
www.example.com
?
user
12.1.3.5
Works fine!
www.example.com
operator
Operator cant see the problem without walking in
the users shoes.
4
Combine Information Across Users
Cluster users to narrow down the root cause
?
www.example.com
Works fine!
www.example.com
5
Consider Recent Network Anomalies
10.0.0.1
Measurement and analysis of routing messages
7081
7018
1239
Route 7018 7081 to 10.0.0.0/8 withdrawn two
minutes ago, making address 10.0.0.1 unreachable!
6
Combine Information From Multiple Domains
10.0.0.1
Typographical error causes prepending of 7018
in path.
Advertisement 7081 7018 automatically discarded
as part of BGP loop prevention
7081
AS 1239 notes suspicious route 7081 7018 from
AS 7081!
7018
?
1239
Operator can notify AS 7081 to fix the problem!
7
Knowledge Plane Implications

• Launch test probes from other vantage points
• Send test traffic from the context of the end
  user
• Log information about DNS server, Web IP address,
  etc.
• Combine information across users
• Consider host pairs (not) able to reach each
  other
• Cluster the host pairs and look for commonalities
• Consider recent anomalies
• Continuously monitor and note possible anomalies
• Preposition results of analysis for future
  reference
• Combine information from multiple domains
• Collect and analyze results at many locations
• Combine information to troubleshoot problems