Bandwidth-Delay Product Explained for Networks Students

Updated November 2026. The bandwidth-delay product (BDP) is a simple idea with an easy unit trap: it is the amount of data that can be “in flight” on a path, obtained by multiplying a link’s bandwidth by the path’s round-trip time (RTT). If a transport window is smaller than that product, the path cannot be filled, even when the advertised megabits look generous. This guide is a bandwidth delay product calculator explanation for networks students, with worked figures you can reproduce on paper. Free Technology Calculators from My Advisers officially includes a tool named Bandwidth–Delay Product, and related official tools named TCP Throughput: Window Limit and Mathis Formula and Sliding Window Link Utilisation. Use those official names; this page teaches the BDP arithmetic those tools exist to support. A wider list of My Advisers tools is in the tools guide. Address math, which is a different networks topic, is in the guide on how to calculate an IPv4 subnet mask and hosts.

The bandwidth-delay product formula, with units you can check

With bandwidth B in bits per second and RTT in seconds:

BDP (bits) = B × RTT

BDP (bytes) = (B × RTT) / 8, using the common eight-bit byte.

If B is given in megabits per second and RTT in milliseconds, convert first: 1 Mbps = 106 bit/s in the SI (decimal) sense used for most link rates, and 1 ms = 0.001 s. Then 100 Mbps × 40 ms = 100 × 106 × 0.040 = 4,000,000 bits = 500,000 bytes. Students often treat “500 KB” loosely; 500,000 bytes is 500 × 103 bytes in decimal kilobytes, or 500,000 / 1,024 ≈ 488.3 KiB in IEC kibibytes. Write the unit. Mixing a decimal megabit link rate with a binary kibibyte window is a common exam slip, not a physical mystery.

RTT here is round-trip time, not one-way delay. A ping-style measurement is a starting point, but a real RTT also includes queueing and processing, and it changes through the day. The illustrations below use labelled RTTs, not official typical values for any Indian ISP or campus network.

Worked examples

Bandwidth (decimal)Labelled RTTBDP in bitsBDP in bytes
100 Mbps40 ms4,000,000500,000
100 Mbps200 ms20,000,0002,500,000
1 Gbps40 ms40,000,0005,000,000
10 Mbps40 ms400,00050,000

The 1 Gbps, 40 ms path has a BDP ten times the 100 Mbps, 40 ms path, as it should: only the bandwidth changed. The 100 Mbps, 200 ms path has a BDP five times the 40 ms path: only the delay changed. A “long fat network” in the textbooks is a path that is both wide and long, so the product is large. A campus link with a small RTT can have a modest BDP even at a high bit rate; a satellite-like RTT at a modest bit rate can still produce a large BDP. This article does not invent a measured satellite RTT. If you need one for an exercise, use the figure your instructor or a measurement gives you, and label it.

Why a TCP window is compared with the BDP

In a sliding-window protocol, the sender may have only a limited amount of unacknowledged data outstanding. If that window, in bytes, is smaller than the BDP of the path, the pipe cannot stay full: the sender will stop and wait for acknowledgements even though the link is idle. A rough upper bound on window-limited throughput is:

Throughput (bits/s) ≈ (window in bytes × 8) / RTT (seconds)

Take a classic 64 KiB window of 65,536 bytes and the labelled 40 ms RTT on a 100 Mbps path. Throughput ≈ 65,536 × 8 / 0.040 = 13,107,200 bit/s ≈ 13.1 Mbps. The link is advertised at 100 Mbps; the window, not the “mega” in the brochure, is the bottleneck in this model. The same window on the 200 ms, 100 Mbps path would be limited to about 2.62 Mbps (65,536 × 8 / 0.200). To fill the 100 Mbps, 40 ms path, the window would need to be at least the 500,000-byte BDP, ignoring overheads and additional congestion control, which this simple bound does not model.

Real TCP also has a congestion window, loss, slow start and path variation. RFC 7323, which updates earlier high-performance TCP extensions, is the official place to read about window scaling, which exists because the original 16-bit TCP window is at most 65,535 bytes and cannot cover a large BDP on its own. This article does not turn RFC 7323 into a configuration tutorial for a particular operating system, and it does not invent kernel-tune values. For window-limited throughput as a named calculation, use the official related FTC tool; do not treat the two-line bound above as the full Mathis formula (that official tool is the place to see the named formula and its assumptions).

A student method that avoids the usual slips

  1. Write B with its unit (Mbps, Gbps) and convert to bits per second with SI powers of ten, unless a question has clearly used another convention.
  2. Write RTT in seconds.
  3. Compute BDP in bits, then divide by 8 for bytes.
  4. If a window is given, convert it to bytes (watch KiB vs KB) and compare it with the BDP, or compute the window-limited throughput and compare it with B.
  5. State what the model left out: loss, overhead, a one-way delay used by mistake, or a bandwidth that was already in bytes per second.

If a question gives distance and a propagation speed instead of RTT, that is a different estimate: a one-way propagation time is distance / speed, and a one-way figure is not an RTT until you have a clearly labelled return path and any extra delay the question includes. Do not treat a straight-line map distance as a measured fibre path.

A second walk-through: filling a 1 Gbps path

Take the 1 Gbps, 40 ms row. B = 109 bit/s, RTT = 0.040 s, BDP = 40,000,000 bits = 5,000,000 bytes. A 65,536-byte window would limit throughput to the same 13.1 Mbps computed earlier, because that bound used only the window and the RTT, not the advertised link rate. On a 1 Gbps path the mismatch is even more obvious: the window-limited rate is a little more than one percent of the advertised rate, in this model. To reach the 5,000,000-byte BDP you need a window that has already used TCP window scaling, in the sense RFC 7323 describes, or you need a delay that is much smaller. If the labelled RTT were 1 ms instead of 40 ms, the same 1 Gbps path would have a BDP of 109 × 0.001 / 8 = 125,000 bytes, and a 64 KiB window would be in the same order of magnitude as the product. The “problem” was not the gigabit label by itself; it was the product of rate and delay.

When two groups of students disagree, they have usually converted RTT as 40 seconds, or used 100 Mbps as 100 × 220 bits, or divided by 8 twice, or compared a bit-count BDP with a byte-count window. Re-do the first example, 100 Mbps and 40 ms, until 500,000 bytes appears without hesitation. Then change one input only.

What BDP does not tell you

BDP does not measure packet loss, it does not size a Wi-Fi radio, and it does not say how long a particular file will take to download once slow start, reverse-path congestion and application pauses are included. It also does not replace utilisation: official related FTC tools address sliding-window utilisation and a named TCP throughput formula, which have their own inputs. If a practical test — iperf or a large transfer — falls well below both the advertised rate and the window-limited bound, look for loss, a smaller effective window, or a slower bottleneck on the path, rather than recomputing BDP until it matches the test. The product is a capacity of the pipe, not a score for the transfer you just ran.

For laboratory write-ups, quote the B, the RTT, the unit of the answer, and whether the window was given in KiB or in decimal kilobytes. That sentence is often worth more marks than a calculator screenshot with no units.

Serialisation delay is not the BDP

A related quantity that appears in the same lecture is serialisation (or transmission) delay: the time to put a packet of S bits onto a link of rate B, which is S/B seconds. For a 12,000-bit packet on a 100 Mbps link that is 0.00012 s, or 0.12 ms. That number is a time, not a volume, and it is usually much smaller than a 40 ms RTT in a wide-area example. Students sometimes add serialisation delay to bandwidth and call the sum a BDP. Keep the definitions apart: serialisation delay tells you how long it takes to place one packet on the wire; RTT tells you how long until an acknowledgement can return; BDP tells you how much data can already have been sent during that round trip at rate B. You may need all three in a longer numerical problem. You should not use one name for another.

If a question gives a one-way propagation delay of 20 ms and asks for BDP, check whether the intended RTT is 40 ms on a symmetric path, or whether the question has already named an RTT. Guessing the missing return path is how two students both write “20 ms” and both lose the factor of two. When the question is silent, state the assumption explicitly: “taking RTT = 2 × 20 ms = 40 ms on a symmetric path, ignoring queueing.”

Keep serialisation units consistent as well: a 1,500-byte Ethernet payload is 12,000 bits only if you mean those 1,500 bytes and no extra headers. If the question includes headers, add them before you compute S/B. A sloppy packet size will not wreck a BDP that never used S, but it will wreck the delay term you add next to the RTT.

When you use the official Bandwidth–Delay Product tool, enter the same converted units you used on paper and compare the byte result with 500,000 for the 100 Mbps, 40 ms example before you trust a more exotic input. If the tool and the paper disagree, the usual cause is an RTT still entered in milliseconds while the formula expected seconds, or a bandwidth still entered in megabits while the field expected bits. Adjust one conversion at a time.

Frequently asked questions

Is BDP the same as throughput?

No. BDP is a volume of data (bits or bytes). Throughput is a rate (bits per second). They are linked when a window and an RTT together limit the rate.

Should I use 1,000 or 1,024 when I convert the result?

Use SI (1,000) for the link rate unless told otherwise, and say whether you then express the byte count in KB or KiB. The official Bandwidth–Delay Product tool will show a figure; your written unit is still your responsibility.

Is ping RTT good enough?

It is a labelled starting point. A transfer’s RTT can be larger than a ping on an idle path. Use the RTT the problem states, or a measurement taken on the path you care about.

Does Free Technology Calculators include this official tool?

Yes. The official November 2026 name list includes “Bandwidth–Delay Product,” plus related official TCP-window tools.

Is this a measurement of my home broadband?

No. It is a networks-class calculation. A speed test measures something else, under conditions the test defines.

Disclaimer

The figures in this article are worked examples at labelled bandwidths and RTTs. They are not measurements of any network, not a tuning guide for a production server, and not a guarantee of throughput. Confirm tool names on the official Free Technology Calculators page, and confirm protocol details in the relevant RFC or your course text.

References

  1. My Advisers, Free Technology Calculators (official “Bandwidth–Delay Product” tool): https://myadvisers.in/free-technology-calculators/.
  2. RFC 7323, TCP Extensions for High Performance: https://www.rfc-editor.org/rfc/rfc7323.
  3. Google Play, Free Technology Calculators: https://play.google.com/store/apps/details?id=com.myadvisers.freetechnologycalculators.

Comments

Popular posts from this blog

Google Ads Campaign Management for Small Businesses in India (2026)

10 Best No-Code Website Builders for Beginners in 2026

Technology Consulting for Small Business Digital Growth in India