As a user, you notice when your application is lagging, and it’s frustrating. As a network engineer, this lag — network latency — is cause for investigation and resolution.
Network latency is how long it takes a packet of data to reach its destination. It is typically measured in round-trip time (RTT) from the sender to the receiver and back, and includes additional delays like propagation delay, processing delay, queuing delay, and encoding delay. Plus, network congestion can add a dynamic element.
- Propagation delay = the time it takes for the head of the packet to reach its destination. Propagation delay is impacted by factors like distance and the mode of transmission.
- Processing delay = once the packet is received, processing delay is the time it takes routers to process the packet header
- Queuing delay = once processed, queuing delay is the time the packet waits in queue before being transmitted
- Encoding delay = preparing the packet for transmission
- Network congestion = caused when the amount of data flowing through (throughput) exceeds its capacity (bandwidth)
However, since networks are dynamic and trips aren’t symmetrical, round-trip time doesn’t let us pinpoint exactly where the delay occurs, nor how long it actually took to get from point A to point B. The more insightful measurement is one-way delay (OWD). One-way delay relies on accurate, synchronized timestamps between the sender and receiver in order to precisely timestamp when the packet departs and arrives.
Because accurate time sync has not been readily available until now, we must rely on RTT and simply divide RTT by 2 in order to approximate one-way measurements. But this crude approximation is not accurate and does not meet the needs of modern applications and use cases. The forward path and reverse path are not the same, and they can also dynamically change over time due to ECMP (inside a single data center) or MPLS over long distance links. We need precise in-network latency numbers to pinpoint where the bottleneck is happening in real time.
With Clockwork, you can accurately measure one-way latency, giving you new insight into your network’s performance, helping you to better manage network congestion, and reducing time to resolution. Clockwork’s time sync solution uses your existing reference clock to synchronize senders and receivers with ns- and us-level sync accuracy, enabling precise timestamps to measure one-way delay. Contact us to learn more and schedule a demo.
Interested in solving challenging engineering problems and building the platform that powers the next generation time-sensitive application? Join our world-class engineering team.