Home
Capacity of the link.
Ping / round-trip latency.
Packet size, default 1460 bytes (TCP over Ethernet).
Bandwidth:
RTT:
MSS:

BDP Quick Reference

Bandwidth-delay product for common link speeds and round-trip times. BDP tells you how much data must be in flight to fully use the link. This is also the minimum TCP window size required for full throughput (before protocol overhead).

Link Speed

What Is the Bandwidth-Delay Product?

The bandwidth-delay product (BDP) is the product of a link's bandwidth and the round-trip time (RTT) across that link. It represents the maximum amount of data that can be "in flight" — sent but not yet acknowledged — on the path at any one moment. Think of it as the capacity of the "pipe" in bytes rather than bits per second.

BDP (bits) = Bandwidth (bps) × RTT (seconds)

A link with 1 Gbps of bandwidth and 100 ms of RTT has a BDP of 1,000,000,000 × 0.1 = 100,000,000 bits, or 12.5 MB. For a sender to fully use that 1 Gbps link, it must keep 12.5 MB of data in flight at all times.

Why BDP Matters

BDP is essential for understanding why high-bandwidth, high-latency links are hard to fill. TCP throughput is limited by the window size:

Max TCP throughput (bps) = TCP Window (bits) ÷ RTT (seconds)

If your window is only 64 KB (the classic TCP default without window scaling), and the RTT is 100 ms, your maximum possible throughput is just:

65,536 bytes × 8 ÷ 0.1 sec = 5.24 Mbps

No matter how fast the link is (100 Mbps, 1 Gbps, 10 Gbps), a single TCP stream on that path cannot exceed ~5.24 Mbps with a 64 KB window. This is why long-distance links — such as satellite, transcontinental fibre, and even fast home internet to overseas servers — often feel slow despite "gigabit" advertised speeds.

TCP Window Scaling to the Rescue

The TCP window scaling option (RFC 7323) allows windows up to 1 GB. Modern operating systems typically enable it by default. To fill a 1 Gbps link at 100 ms RTT, you need 12.5 MB of window. To fill 10 Gbps at the same RTT you need 125 MB — which is at the edge of what most OS stacks support per socket.

The practical consequences:

  • LAN (RTT < 1 ms): small windows work fine; BDP is tiny.
  • Metro / regional (10–50 ms): moderate windows needed; most modern stacks handle this easily.
  • Transcontinental (80–150 ms): large windows required; window scaling is essential.
  • Satellite (500–700 ms): "long fat network" — extremely large windows, plus performance-enhancing proxies (PEPs) are typically needed.

Long Fat Networks (LFNs)

A link with a BDP greater than about 64 KB is called a long fat network or "LFN" (and sometimes "long fat pipe" or LFP). Classic TCP without window scaling cannot fill an LFN with a single stream — you need either window scaling, multiple parallel connections, or both. This is why download managers split files into many parallel streams: each stream gets its own window, effectively multiplying the throughput.

Bandwidth-Delay Product vs. Throughput

The BDP is not the maximum amount of data you can transfer — it's the amount of data currently in transit. Throughput is a rate (bits per second); BDP is a volume (bits, or bytes). They are related via RTT:

Throughput (bps) = BDP (bits) ÷ RTT (seconds)

When you measure a link with a single TCP stream, the highest rate you will see is the window size divided by the RTT — and it cannot exceed the link bandwidth. If the window is at least BDP, you fill the link. If it's less, you're window-limited.

Practical Example: Satellite vs Fibre

Satellite link — 100 Mbps, 600 ms RTT

BDP = 100,000,000 × 0.6 = 60,000,000 bits = 7.5 MB

A single TCP stream with a 64 KB window can only reach 0.87 Mbps — less than 1% of the link speed.

Transatlantic fibre — 1 Gbps, 100 ms RTT

BDP = 1,000,000,000 × 0.1 = 100,000,000 bits = 12.5 MB

A single TCP stream with a 64 KB window reaches only 5.24 Mbps — about 0.5% of the link speed. Window scaling to 16 MB+ is essential.

How to Use This Calculator

  • BDP Calculator: Enter link bandwidth and RTT to get BDP in bits, bytes, KB, MB, and packets.
  • Max Throughput: Enter your TCP window and RTT to see the maximum single-stream throughput. Optionally enter link bandwidth to see what percentage of the link you can fill.
  • Required Window: Enter target bandwidth and RTT to see the minimum TCP window you need for full-speed single-stream transfer.

❓ Bandwidth-Delay Product FAQ

What is the bandwidth-delay product in simple terms?

It's the amount of data that must be "in flight" on a network at any given moment to fully use the link. If a link is 1 Gbps and RTT is 100 ms, then 12.5 MB of data is either travelling in one direction or coming back as an ACK at any instant. To saturate the link, the sender must have at least that much data in flight.

What is the formula for bandwidth-delay product?

BDP = Bandwidth × RTT. Keep units consistent: bandwidth in bits per second and RTT in seconds gives BDP in bits. Divide by 8 to get bytes; divide by 1,000,000 to get MB (or by 1,048,576 if using binary MB).

Why is a 64 KB TCP window not enough on long-distance links?

The classic TCP window of 65,535 bytes only allows ~5.24 Mbps throughput at 100 ms RTT. On a 1 Gbps transatlantic link, that's under 1% of the available bandwidth. TCP window scaling (RFC 1323 / RFC 7323) exists specifically to solve this — it allows windows up to 1 GB.

What is a "long fat network" (LFN)?

A network path whose bandwidth-delay product exceeds about 64 KB. On an LFN, classic TCP without window scaling cannot fill the link, because the receiver would have to "pause" the sender every 64 KB until ACKs come back — leaving the wire idle during the wait. Satellites, transcontinental fibre, and even fast home connections to distant servers are LFNs.

Does BDP apply to UDP?

Not the same way. UDP has no windowing or congestion control — the sender can fire packets as fast as it likes. But the BDP still describes the pipe capacity — how much data can be in transit — so it's useful for sizing UDP-based protocols (VoIP, QUIC, some VPNs) and for understanding buffer bloat.

How does BDP relate to buffer sizing in routers?

Router buffers should hold roughly one BDP of data to keep the pipe full during bursts and avoid drops. The famous "rule of thumb" is buffer ≈ BDP, though modern research suggests BDP / √N is enough when many flows share a link (where N is the number of flows). Too little buffer and the link under-utilises; too much and you get buffer bloat.

Why is my download speed slower than my bandwidth on long-distance transfers?

Almost always a BDP/window issue. If your TCP window (or the server's) is too small relative to the BDP, the single stream can't fill the link. Solutions: larger windows (usually automatic with modern OSes), more parallel connections (download managers, browsers), or CDNs that keep content closer (lower RTT) to you.

What is MSS and why does the calculator ask for it?

Maximum Segment Size (MSS) is the largest TCP payload per packet. For standard Ethernet (1500 byte MTU) it's 1460 bytes; for IPv4 the minimum is 536; for jumbo frames it's 8960. Dividing BDP by MSS gives you the number of packets that must be in flight to fill the link. Useful for tuning driver buffers and understanding round-trip counts.

How do I increase the TCP window on my system?

On Linux, tune net.ipv4.tcp_rmem and net.ipv4.tcp_wmem. On Windows, the receive window auto-tuning is on by default; the relevant knob is netsh int tcp set global autotuninglevel=normal. On macOS the window scales automatically. Most modern stacks scale up well by default; only very high-BDP paths need manual tuning.

Is this calculator free?

Yes. This tool is completely free, requires no registration, and runs entirely in your browser. Your inputs are never sent to a server.