Performance Metrics for All

Morton

ip network performance metrics measurement instrumentation

@article{morton:ieee-ic2009,
  title={Performance Metrics for All},
  author={Morton, A.},
  journal={{IEEE} Internet Computing},
  volume={13},
  number={4},
  pages={82--86},
  year={2009}
}

Performance metrics should

• Be unambiguously specified and named
• Have clear options and units of measure
• Characterize relevant performance aspects
• Lead to implementation

Three steps in defining metrics/measurements:

• Define boundaries and interfaces
  • Interfaces between areas of responsibility
  • Most likely place for measurement points
• Identify reference events at the boundaries
• Define performance metrics based on reference events at the boundaries

Performance standards often broken into three components:

• Metric definitions
• Measurement methods
• Numerical performance objectives

Notable requirements of different users may include:

• Suitability of passive live measurements, or active measurement with synthetic traffic
• Laboratory or deployed environment
• Accuracy
• Access to measurement points and configuration information
• System scale
• On-demand or continuous measurement
• Reporting formats
• May help determine sufficient coverage of metrics via ANSI X3.102 styled 3x3 matrix
• Cover each area
• Follow up on this

IPPM metrics:

• Connectivtiy
• One-way delay and loss
• Round-trip delay
• Delay variation
• Loss patterns
• Packet reordering
• Bulk transport capacity
• Link bandwidth capacity
• Packet duplication

Of note:

• IETF Benchmarking Methodology Working Group