Why slow on the first and second HTTP request?

[campos]

Hi all,

I am trying to use javascript to send an HTTP/post request to a server
multiple times continously. But the time used on the 1st and 2nd files
are longer than others. My idea is that because of the 3-way hand
shake, time to be used on transferring the first file shoule be 1 RTT
longer than that of others. But the results are different from what
I'm thinking of. I captured the packets via Wirshark and found out
that at the beginning of transferring the 1st file, the server waited
for the ACK from client side after sending out 2920 bytes. But on
looking at the TCP header, I found that all of the TCP packet claimed
that the window size is 65535. I wonder why the 1st and 2nd files need
much more time.

Thanks in advance.

Below are the http request files:
Response content size = 60 KB count(1): The time = 1342 milliseconds.
Response content size = 60 KB count(2): The time = 351 milliseconds.
Response content size = 60 KB count(3): The time = 190 milliseconds.
Response content size = 60 KB count(4): The time = 180 milliseconds.
Response content size = 60 KB count(5): The time = 191 milliseconds.
Response content size = 60 KB count(6): The time = 190 milliseconds.
Response content size = 60 KB count(7): The time = 190 milliseconds.
Response content size = 60 KB count(8): The time = 191 milliseconds.
Response content size = 60 KB count(9): The time = 190 milliseconds.
Response content size = 60 KB count(10): The time = 180 milliseconds.

 

[Doug Miller]

I am trying to use javascript to send an HTTP/post request to a server
multiple times continously. But the time used on the 1st and 2nd files
are longer than others. [snip]
Below are the http request files:
Response content size = 60 KB count(1): The time = 1342 milliseconds.
Response content size = 60 KB count(2): The time = 351 milliseconds.
Response content size = 60 KB count(3): The time = 190 milliseconds.

[snip remaining timings similar to the last one]

Google "statistical multiplexing" or "dynamic bandwidth allocation" for one
possible explanation.