Optimal TCP Window Size For Long Distance WAN Link

If you have a requirement to copy large amounts of data along way around the world you may find that despite your link being 60Mb/s if its 5,000 miles away you only can transfer files at much less like 10Mb/s. The cause of this is generally the TCP Window Size is optimized by OS and FTP clients by default to work on networks with less distance and less network round trip latency.

When using TCP to transfer data the two most important factors are the TCP window size and the round trip latency. If you know the TCP window size and the round trip latency you can calculate the maximum possible throughput of a data transfer between two hosts, regardless of how much bandwidth you have.

TCP-Window-Size-in-bits / Latency-in-seconds = Bits-per-second-throughput

The standard Windows TCP window is 64KB

64KB = 65536 Bytes = 524288 bits

To calculate the round trip delay time can be done in many ways, the easiest is to use ping.exe as this records RTD in ms.

In the above the average latency is 38ms. So using formula we can calculate the theoretical maximum.

524288 / 0.038 = 13797053 bp/s = 13.8 Mbp/s

So you might want to know what you can do to make it faster? As with any equation the answer is adjust part so either increase the TCP window size and/or reduce latency. In many ways the latency is not something we can control easily but we can easily adjust the TCP window size. If we know the latency and we know the size of pipe we have we should be able to calculate the best TCP window size.

(Link speed * Latency / 8) = TCP Window size Bytes

So if we have a 60Mb/s link across a route 5,042 miles incurring latency of 173ms.

(60000000*0.173)/8 = 1297500

The downside to increasing the TCP window size on your server is that these larger packets increase the buffer size, this buffer sits in memory. Also if larger packets are lost, then more data is lost and therefore more needs retransmitting.