On 5/15/06, John Heffner <jheffner@xxxxxxx> wrote:
Constantinos Makassikis wrote: > As it can be seen, we hardly reach 70 % of the value predicted by the > formula > and apparently it seems that it is due to the fact that MAX SND > remains relatively > low compared to SNDBUF. This is due to the fact that (differently than BSD) Linux treats socket buffer sizes as limits on the amount of kernel memory that socket buffers can consume, rather than the number of in-band TCP data they can hold. There's a certain amount of overhead involved in storing the buffered data, so you'll see a correspondingly smaller window size, if the TCP window is limited by the socket buffer size. -John
Thanks for pointing this significant conceptual difference. Is there a way to predict the amount of overhead or the in-band TCP data that can be hold ? I believe that the reason the W/RTT formula is verified so poorly lies in the fact I don't have a good estimation of W. Constantinos - : send the line "unsubscribe linux-net" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
