Latency Budget Calculator
Inputs
| Propagation delay | 20 ms |
|---|---|
| Serialization delay | 1 ms |
| Queuing delay | 5 ms |
| Processing delay | 2 ms |
Latency Budget Calculator
Break network latency into its four components — propagation, serialization, queuing, and processing delay — to find one-way delay and round-trip time. Useful for diagnosing where a connection spends its time.
Inputs
Delay Components
Results
Enter a value to see results.
Round-Trip Time
One-Way Delay
Latency Budget
Network latency is rarely a single number. The delay a packet experiences from one end of a path to the other is the sum of several distinct effects, each with its own cause and its own cure. A latency budget treats total delay as a sum of these parts, so you can see where the time is spent and which component to attack when an application feels sluggish.
The four components
One-way delay is the sum of four classic components, then round-trip time is twice that for a symmetric path:
t1=tp+ts+tq+td RTT=2×t1where is propagation delay, is serialization delay, is queuing delay, and is processing delay.
- Propagation delay (). The time for the signal to cross the physical distance, governed by the speed of light in the medium. Light in optical fibre travels at roughly two-thirds of its vacuum speed, so a path adds about 5 ms of propagation per 1,000 km each way. Distance is the only lever.
- Serialization delay (). The time to push every bit of a packet onto the wire, equal to packet size divided by link rate. A 1,500-byte frame on a 1 Gbps link serializes in about 12 microseconds. Faster links and smaller packets shrink it.
- Queuing delay (). The time a packet waits in a buffer behind other traffic. It is near zero on an idle link and climbs steeply as utilisation approaches 100%. This is the most variable component and the usual cause of jitter.
- Processing delay (). The time each device spends parsing headers, looking up routes, and forwarding. On modern hardware it is small and steady, typically microseconds per hop.
Worked example
Consider a path with 20 ms of propagation, 1 ms of serialization, 5 ms of queuing, and 2 ms of processing delay. The one-way delay is:
t1=20+1+5+2=28 msThe round-trip time is twice that:
RTT=2×28=56 msSo a ping across this path would read about 56 ms, and a one-directional stream would see 28 ms of latency. Propagation dominates here, which is typical of a medium-distance internet path.
Which component dominates
The mix changes with the situation, and that tells you where to spend effort.
| Situation | Dominant component |
|---|---|
| Long-distance / intercontinental link | Propagation |
| Saturated or congested link | Queuing |
| Slow last-mile link, large packets | Serialization |
| Many hops, deep inspection | Processing |
On a transcontinental link, propagation can be tens of milliseconds and swamps everything else, so no amount of faster hardware helps — only moving the endpoints closer does. On a congested link, queuing can spike into the hundreds of milliseconds, and adding capacity or active queue management is the fix.
Why round-trip time matters
Round-trip time is the figure most protocols actually care about. A request must reach the destination and the reply must come back, so for a symmetric route the round trip covers the one-way distance twice — hence the factor of two. TCP cannot acknowledge data faster than one RTT, which means high-latency links throttle throughput even when bandwidth is abundant. That coupling between latency and throughput is why long, fast links need large send windows, and it is the link between this calculator and bandwidth-delay sizing.
Real paths are not always perfectly symmetric: the forward and return routes can differ, so a measured ping may not be exactly double a one-way measurement. The factor of two is a sound planning assumption, not a guarantee.
Reducing the budget
Once you know which component dominates, the remedy follows. Propagation responds only to shorter routes, which in practice means a content delivery network or edge presence that places servers near users. Queuing responds to more capacity, traffic shaping, and queue management that keeps buffers from filling. Serialization responds to faster links and smaller frames. Processing is usually already small but benefits from hardware forwarding on busy devices. Measuring each piece first ensures you pull the lever that actually moves the total.
Related calculators
Once you know the round-trip time, size the in-flight data and TCP window with the Bandwidth-Delay Product Calculator calculator. To see how latency caps throughput on a lossy link, use the TCP Throughput Calculator calculator. To estimate how long a bulk transfer takes end to end, see the Data Transfer Time Calculator calculator.
Frequently Asked Questions (FAQ)
What is a latency budget?
A latency budget is a breakdown of total network delay into the parts that cause it, so you can see where the time goes and how much room is left before an application becomes unresponsive. Network one-way delay has four classic components: propagation (distance), serialization (clocking bits onto the wire), queuing (waiting in buffers), and processing (device handling).
Adding them gives the one-way delay; doubling it gives the round-trip time. Treating latency as a budget makes it clear which component to attack: a long-distance link is dominated by propagation, while a congested link is dominated by queuing.
What is the difference between propagation and processing delay?
Propagation delay is purely a function of distance and the medium: light in fibre travels at roughly two-thirds the speed of light in a vacuum, so a 3,000 km path adds about 15 ms each way no matter how fast the equipment is.
Processing delay is the time each router or switch spends inspecting and forwarding a packet, typically tens of microseconds on modern hardware. Propagation usually dominates on long-haul links, while processing only becomes significant when a packet crosses many hops or passes through deep packet inspection.
Why is round-trip time twice the one-way delay?
A request has to travel to the destination and the response has to travel back, so for a symmetric path the round trip covers the one-way distance twice. That is why RTT equals two times the one-way delay in this calculator.
Real paths are not always symmetric — the forward and return routes can differ — so a measured ping may not be exactly double a one-way measurement. RTT matters because protocols like TCP cannot acknowledge data faster than one round trip, which caps throughput on high-latency links.
How can I reduce latency?
Target the component that dominates your budget. To cut propagation delay, move servers closer to users with a CDN or edge location, since you cannot beat the speed of light.
To cut queuing delay, add capacity, apply traffic shaping, or enable active queue management so buffers do not fill. To cut serialization delay, use faster links or smaller packets. Processing delay is usually already small, but offloading and hardware forwarding help on busy devices. Measuring each component first tells you which lever will actually move the number.