Exit Zen
111
../../protocols/tcp.rst:334 • ../../protocols/tcp.rst:347
English
1 msec
0/100 · 6
112
../../protocols/tcp.rst:334
English
524 Mbps
0/100 · 8
113
../../protocols/tcp.rst:335 • ../../protocols/tcp.rst:348
English
10 msec
0/100 · 7
114
../../protocols/tcp.rst:335
English
52.4 Mbps
0/100 · 9
115
../../protocols/tcp.rst:336 • ../../protocols/tcp.rst:349
English
100 msec
0/100 · 8
116
../../protocols/tcp.rst:336
English
5.24 Mbps
0/100 · 9
117
../../protocols/tcp.rst:337 • ../../protocols/tcp.rst:350
English
500 msec
0/100 · 8
118
../../protocols/tcp.rst:337
English
1.05 Mbps
0/100 · 9
119
../../protocols/tcp.rst:340
English
To solve this problem, a backward compatible extension that allows TCP to use larger receive windows was proposed in :rfc:`1323`. Today, most TCP implementations support this option. The basic idea is that instead of storing `snd.wnd` and `rcv.wnd` as 16 bits integers in the :term:`TCB`, they should be stored as 32 bits integers. As the TCP segment header only contains 16 bits to place the window field, it is impossible to copy the value of `snd.wnd` in each sent TCP segment. Instead the header contains `snd.wnd >> S` where `S` is the scaling factor ( :math:`0 \le S \le 14`) negotiated during connection establishment. The client adds its proposed scaling factor as a TCP option in the `SYN` segment. If the server supports :rfc:`1323`, it places in the `SYN+ACK` segment the scaling factor that it uses when advertising its own receive window. The local and remote scaling factors are included in the :term:`TCB`. If the server does not support :rfc:`1323`, it ignores the received option and no scaling is applied.
0/10230 · 1023
120
../../protocols/tcp.rst:342
English
By using the window scaling extensions defined in :rfc:`1323`, TCP implementations can use a receive buffer of up to 1 GByte. With such a receive buffer, the maximum throughput that can be achieved by a single TCP connection becomes :
0/2340 · 234
121
../../protocols/tcp.rst:347
English
8590 Gbps
0/100 · 9
122
../../protocols/tcp.rst:348
English
859 Gbps
0/100 · 8
123
../../protocols/tcp.rst:349
English
86 Gbps
0/100 · 7
124
../../protocols/tcp.rst:350
English
17 Gbps
0/100 · 7
125
../../protocols/tcp.rst:357
English
These throughputs are acceptable in today's networks. However, there are already servers having 10 Gbps interfaces... Early TCP implementations had fixed receiving and sending buffers [#ftcphosts]_. Today's high performance implementations are able to automatically adjust the size of the sending and receiving buffer to better support high bandwidth flows [SMM1998]_.
0/3680 · 368
126
../../protocols/tcp.rst:362
English
TCP's retransmission timeout
0/280 · 28
127
../../protocols/tcp.rst:364
English
In a go-back-n transport protocol such as TCP, the retransmission timeout must be correctly set in order to achieve good performance. On one hand, if the retransmission timeout expires too early, then bandwidth is wasted by retransmitting segments that have already been correctly received. On the other hand, if the retransmission timeout expires too late, then bandwidth is wasted because the sender is idle waiting for the expiration of its retransmission timeout.
0/4670 · 467
128
../../protocols/tcp.rst:366
English
A good setting of the retransmission timeout clearly depends on an accurate estimation of the round-trip-time of each TCP connection. The round-trip-time differs between TCP connections, but may also change during the lifetime of a single connection. For example, the figure below shows the evolution of the round-trip-time between two hosts during a period of 45 seconds.
0/3730 · 373
129
../../protocols/tcp.rst:372
English
Evolution of the round-trip-time between two hosts
0/500 · 50
130
../../protocols/tcp.rst:375
English
The easiest solution to measure the round-trip-time on a TCP connection is to measure the delay between the transmission of a data segment and the reception of a corresponding acknowledgment [#frttmes]_. As illustrated in the figure below, this measurement works well when there are no segment losses.
0/3010 · 301