Network performance monitoring in iPM spans two related but distinct things: how much traffic is flowing, and how healthy the connection actually is — a link can be far from saturated and still be performing badly.
What's measured
- Throughput — inbound and outbound traffic per interface, typically shown as both a current rate and a graph over time
- Errors and discards — packets an interface failed to send or receive cleanly; a rising error count usually points to a physical-layer problem (a bad cable, a failing transceiver, a duplex mismatch) rather than a capacity problem
- Latency — round-trip time to a host, typically via ping-based checks; a sudden increase often signals congestion or a routing change somewhere along the path
- Packet loss — missed responses to ping checks, one of the clearest signs of an unhealthy link even when throughput looks otherwise normal
Interpreting throughput against capacity
A raw throughput number means little without context — 500 Mbps is comfortable headroom on a 10 Gbps link and effectively saturated on a 1 Gbps one. Where possible, alert on utilization as a percentage of the interface's known capacity rather than on an absolute throughput figure, so the same threshold logic works sensibly across links of very different sizes.
Errors matter even at low traffic
Unlike throughput, a nonzero and rising error or discard count is worth investigating regardless of how busy the link currently is — it's rarely caused by volume alone, and tends to get worse (and harder to diagnose after the fact) the longer it's left unaddressed.
For deeper flow-level analysis — identifying exactly which conversations are consuming bandwidth — see the Netflow capability, which complements interface-level throughput with per-conversation detail.