Measurements for Network Operations

Part 1: Outline

Characteristics of the Internet

Curse of the Dumb Network

Traffic Measurement: Science vs. Engineering

Detecting Performance Problems

Network Operations: Excess Traffic

Network Operations: Denial-of-Service Attack

Network Operations: Link Failure

Network Operations: Measure, Model, and Control

Network Operations: Time Scales

Terminology: Measurements vs Metrics

Collection of Measurement Data

Controlling Measurement Overhead

Reducing Packet/Flow Measurement Overhead

Basic Properties

Simple Network Management Protocol (SNMP)

SNMP: Management Information Base (MIB)

SNMP: Applications of SNMP Traffic Statistics

SNMP: Measurement Limitations

SNMP: Conclusions

Packet Monitoring

Packet Monitoring: Tapping the WAN Link

Packet Monitoring: Selecting the Traffic

Extracting Data from IP Packets

Packet Monitoring: IP Header Traces

Packet Monitoring: TCP Header Analysis

Packet Monitoring: Packet Contents

Packet Monitoring: Monitor at the Edge

Packet Monitoring: Monitor in the Core

Packet Monitoring: System Constraints

Packet Monitoring: Packet Sampling

Packet Monitoring: Application-Layer Sampling

Packet Monitoring: PSAMP IETF Activity

Packet Monitoring: Conclusions

Flow Measurement: Outline

Flow Measurement: IP Flows

Flow Measurement: Recording Traffic Statistics

Flow Measurement: Versus Packet Monitoring

Flow Measurement: Tapping the Link

Flow Measurement: Mechanics of Flow Cache

Flow Measurement: Evicting Cache Entries

Flow Measurement: Packet Sampling

Flow Measurement: Flow-Level Sampling

Flow Measurement: Aggregation

IETF Standards Activity

Flow Measurement: Conclusions

Part 2: Outline

Traffic Representations for Network Operators

End-to-End Traffic & Demand Models

Domain-Wide Network Traffic Models

Path Matrix: Operational Uses

Traffic Matrix: Operational Uses

Demand Matrix: Motivating Example

Coupling of Inter and Intradomain Routing

Intradomain Routing: “Hot Potato”

Demand Model: Operational Uses

Populating the Domain-Wide Models

Inference: Network Tomography

Tomography: Formalizing the Problem

Tomography: Single Observation is Insufficient

Tomography: Challenges

Promising Extension: Gravity Models

Mapping: Remove Traffic Assumptions

Traffic Mapping: Ingress Measurement

Traffic Mapping: Egress Point(s)

Traffic Mapping: Combining the Data

Mapping: Challenges

Direct Observation: Overcoming Uncertainty

Direct Observation: Straw-Man Approaches

Direct Observation: Trajectory Sampling

Trajectory Sampling: Fields Included in Hashes

Trajectory Sampling

Trajectory Sampling: Summary

Example Application: Intradomain Traffic Engineering

Traffic Engineering by Tuning Link Weights

Weight Optimization

Incorporating Operational Realities

Conclusions

Interesting Research Problems