Jitter Buffer Packet Losses
Understand jitter buffer packet losses — packets declared lost by the buffer after waiting past their deadline, and how they differ from network-level loss.
Jitter buffer packet losses is a cumulative count of packets that the jitter buffer has declared permanently lost. After waiting for a packet past its maximum deadline, the buffer concludes the packet will never arrive and marks it as lost. Each lost packet represents a gap in the media stream that cannot be recovered.
Think of it as waiting for a delivery that never shows up. After enough time passes, you stop waiting and move on — the package is declared lost.
How It Works
When the jitter buffer detects a gap in the sequence numbers of incoming packets, it begins a waiting period. During this window, the missing packet might still arrive (out of order or retransmitted). If the packet does not arrive before the buffer needs to advance playback past that point, it is declared lost.
This is different from network-level packet loss measured by RTCP:
- Network-level loss counts packets that never arrived at the receiver at all, based on sequence number gaps in RTCP Receiver Reports.
- Jitter buffer loss counts packets the buffer gave up on — which includes network-level losses but also accounts for packets that arrived so late they were already past the buffer's patience window.
The jitter buffer loss count is cumulative and increases monotonically throughout the stream.
Every loss is audible or visible
Unlike network losses that might be recovered by retransmission (NACK/RTX), jitter buffer losses are final. The buffer has already passed the playback point for these packets. For audio, each loss creates a gap or artifact. For video, losses contribute to frame corruption, artifacts, or freezes.
Why It Matters
Jitter buffer packet losses are the most accurate indicator of actual media quality degradation at the receiver. While network-level loss tells you what happened on the wire, buffer-level loss tells you what the user actually experienced. A system with 5% network loss but effective retransmission might show near-zero buffer losses. Conversely, a system with only 1% network loss but poor retransmission might show significant buffer losses.
For SIP load testing, comparing network-level and buffer-level losses reveals how effective your error recovery mechanisms are. If buffer losses are much lower than network losses, retransmission (RTX) is working well. If they are similar, recovery is failing.
Thresholds
| Level | Value (cumulative per stream) | Interpretation |
|---|---|---|
| Good | 0 - 10 | Minimal unrecoverable loss |
| Warning | 10 - 50 | Notable quality degradation, recovery insufficient |
| Critical | Above 50 | Severe quality impact, many gaps in media |
Common Causes
| Cause | Explanation |
|---|---|
| Network packet loss | Packets dropped on the wire and never arriving |
| Failed retransmission | RTX requests sent but responses not arriving in time |
| Severe jitter | Packets arriving so late the buffer has given up |
| Firewall drops | Stateful firewalls dropping packets mid-stream |
| Congestion-induced loss | Network equipment dropping packets under overload |
How to Fix It
- Address network-level loss — Buffer losses start with network losses. Follow the guidance in Packet Loss to reduce drops on the network path.
- Improve retransmission success — Check the RTX Requests and RTX Success metrics. If requests are high but success is low, the retransmission path has problems.
- Increase buffer capacity — A larger buffer gives more time for late and retransmitted packets to arrive. Verify the buffer's maximum delay setting accommodates your network conditions.
- Compare with network loss — If buffer losses match network losses closely, retransmission is not helping. If buffer losses are much lower, retransmission is effective and the remaining losses are unrecoverable.
- Check for concentrated bursts — A few large burst losses are harder to recover than many scattered single-packet losses. Use time-series analysis to identify burst patterns.
Related Metrics
- Packet Loss — Network-level loss from RTCP, for comparison
- Late Packets — Packets that arrived but past deadline (late is distinct from lost)
- RTX Requests — Retransmission attempts to recover missing packets
- RTX Success — Successfully recovered packets, reducing the final loss count
- Packets Pushed — Packets successfully delivered to the decoder
Jitter Buffer Average Jitter
Understand the jitter buffer's internal average jitter measurement — how it differs from network-level jitter and why it provides a more accurate picture.
Jitter Buffer Duplicates
Understand jitter buffer duplicate packets — what causes them, when they matter, and how they relate to retransmission and network behavior.