Jitter

The variation in latency from packet to packet.

Definitions only — a term explained here is not a service we sell.

Jitter is the variation in delay between packets on the same path; for IP it is standardized as packet delay variation (ITU-T Y.1540, IETF RFC 3393). A stream with a stable average but a wide spread arrives unevenly, and real-time applications feel that spread.

Why it matters

Voice and video buffer to absorb jitter. When the variation exceeds the buffer, the application drops or delays packets, so a path with a good average can still sound bad.

A delay distribution

A delay distribution
Value Zone What it means
p50 40 ms Typical packet
p95 180 ms Slow tail
p95 − p50 140 ms Variation

Illustrative example; not a measurement of our network.

The median and the slow tail describe jitter better than one average. The values are illustrative.
Text description

The strip plots two points from a delay distribution. The spread between the typical packet and the slow tail is one illustrative way to read delay variation; it is not the definition. A single average hides that spread.

What to check

Ask for percentiles over a window, not one average. Measure at the application's packet rate, because jitter on a ping is not jitter on a voice stream. The latency term and the packet loss term cover the related failures.

FAQ

Is jitter the same as latency?

No. Latency is the delay; jitter is how much that delay changes from packet to packet.

What causes jitter?

Queueing, route changes, contention and scheduling. A congested link usually shows both higher latency and wider jitter.

How is jitter measured?

As the variation in one-way delay or round-trip time over a window. The method and the window change the number.

Learn more

Sources

  1. ITU-T Y.1540: IP packet transfer and availability performance parameters (IPDV)
  2. IETF RFC 3393: IP Packet Delay Variation Metric for IP Performance Metrics (IPPM)