Why Latency, Packet Loss, and Jitter Cripple Your Apps—and How to Tackle Them

Latency

Early network protocols were designed for LANs where round‑trip times were only a few milliseconds, but on the global Internet a single round‑trip can exceed 200 ms. Queueing theory shows that as link utilization approaches capacity, waiting time grows dramatically; a 10 Gbps link at 99.9 % utilization can cause severe queuing delay. TCP combats high latency by keeping enough data in flight, avoiding the inefficiency of waiting for an ACK after each packet, which would limit a 200 ms RTT link to just five packets per second. However, short‑lived connections suffer from TCP slow‑start, and frequent connection churn can dramatically degrade performance, especially when DNS lookups add additional delay.

Packet Loss

In real networks, bits can be corrupted or interfered with, especially over wireless links, leading to packet loss. When a packet is lost, TCP must retransmit after detecting the gap, which can stall the pipeline. For example, on a 200 ms RTT link sending 1 000 packets per second, losing packet 500 means the sender continues up to packet 700 while the receiver waits for the missing packet to be retransmitted, causing out‑of‑order delivery and additional buffering. High loss combined with high latency can overwhelm TCP buffers, forcing the connection to pause until the missing data is recovered. Excessive sending rates can also cause routers to drop packets, triggering fast retransmit but potentially leading to another slow‑start cycle that hurts short web requests.

Jitter

Jitter is the variability of packet delivery time caused by fluctuating queue lengths in routers. While TCP can tolerate some jitter by using conservative RTT estimates and longer timeouts, real‑time audio and video are highly sensitive; large jitter results in stutter, frame loss, or audio‑video desynchronization, forcing applications to either increase latency buffers or accept higher loss rates.

Conclusion

Choosing a network path with low latency, low packet loss, and low jitter is critical for optimal application performance. In multi‑path environments, intelligent routing algorithms—such as those employed by Noction—can dynamically select the best route based on real‑time loss and latency metrics, thereby improving throughput and reducing cost.