There\u2019s a tool called iPerf which is essentially a CLI speed test server and client. It runs better on Linux because there\u2019s less shenanigans that go on inside the network stack but works well on Windows as well.<\/p>
A few things you need to be aware of before running this if you want accurate results.<\/p>
The first is a thing called Bandwidth Delay Product (BDP). In the network stack all applications request an amount of memory to be allocated as a buffer for incoming data. If you make this too small; you\u2019re going to artificially limit your bandwidth.<\/p>
To calculate BDP it\u2019s a little complicated \u2013 there\u2019s a calculator here\u00a0https:\/\/www.speedguide.net\/bdp.php<\/a><\/p> 1. Determine MSS, maximum anticipated RTT latency and advertised maximum bandwidth. If you\u2019re running iPerf and doing a UDP based test (remembering UDP has a much lower overhead than TCP so will give you a more accurate result) you need to tell iPerf the amount of bandwidth you\u2019re expecting the link to support with the -b switch<\/p> Also keep in mind if you\u2019re running this on a live environment during production hours you\u2019re going to saturate a link.<\/p> The last is windows firewall is not your friend here.<\/p> Fortigate also has a poor mans version of iPerf but you can\u2019t start it in server mode.<\/p>
(for example, 1460 MSS, 300ms max latency, 6Mbit\/s max bandwidth).
2. Find the unscaled RWIN value (largest even multiple of MSS less than 65535):
65535 \/ 1460 (MSS) = 44.9
round down to even number = 44
44 x 1460 = 64240 (this is the optimal\u00a0unscaled<\/u>\u00a0RWIN value for broadband connections)
3. Multiply the unscaled RWIN by 2 until it is larger than BDP
Our BDP for 6Mbps @ 300 latency is:
6000 kbps x 300ms = 1800000 \/ 8 = 225000
64240 (unscaled RWIN) x 2 x 2 … =\u00a0256960<\/b>\u00a0(number larger than BDP, this is the optimal RWIN)<\/p>