Root Cause Analysis

1
Root Cause Analysis

• Abrar Ul Haque
• Jonathan D'Andries

2
Root Cause Analysis

• What is it about ?
• Definition
• Finding the root-cause of a problem from among a
  variety of symptoms
• Example

3
Downstream Suppression

• Mostly being done in the commercial products
• Management system polls the network entities
• Reduces the number of alarms
• Ignores failures from devices downstream

4
Works well in some cases

• Suppose S1 fails

NMS
Router 1
Router 4
Router 2
Router 3
Switch 3
Switch 1
Switch 2
H 1
H 7
H 2
H 3
H 4
H 5
H 6
5
But
NMS

Router 1
Router 2
Router 4
Router 3
6

• Several Drawbacks
• Only works with hierarchical networks
• Cannot identify a problem unless there is an
  event directly associated with it.
• All symptoms do not propagate downstream
• NOT Root Cause Analysis

7
Correlation

• Finding relationships between symptoms to
  determine root cause
• RCA products classified according to their
  correlation techniques

8
Correlation (cont..)

• Rule-based Reasoning
• Network specific rules to analyze events

9
Correlation (cont..)

• Model-Based Reasoning
• Model individual components and the collaboration
  among components.

10
Correlation (cont..)
11
Correlation (cont..)

• State Transition Graphs
• State machines - token, state, arc, action

Ping Hub
Ping again
Response
ping
Response
No response
Response
Receive response
Ring router
Receive response
No response
Request
No response
Response received from router Send alarm
Ping again
12
Correlation (cont..)

• Codebook Correlation
• Matching symptoms to problem signatures

13
Correlation (cont..)

• Case-based Reasoning
• Use the history of problem solutions to address
  current problems

Case Library
Process
Adapt
Retrieve
Input
14
RCA products

• Netcool
• Rule-based Reasoning
• inCharge
• Codebook Correlation
• NerveCenter
• State Transition Graphs
• NetXamine
• CMIP based

15
Questions

• Abrar (abrar_at_cc)
• Jonathan (bane_at_cc)