-2

Consider an Access Point with a packet queue that can hold up to 50 packets and a network whose bottleneck is a link with a datarate of 50 Mbit/s.

There are quite a few nodes connecting wirelessly to the Access Point, one of them uploads a file (1.5 GB) via FTP to a remote server. When observing the average throughput of the FTP upload, I have noticed it is around 13.7 Mbit/s. And now I wonder how this is possible? The bottleneck is 50 Mbit/s, right?

The only reason that I can think of is this: The bottleneck must be somewhere else. And since all physical connections of the network support a higher datarate, it must be with the Access Point.

The TCP MSS = 1460 bytes, its advertised receive window is scaled to 1,460,000 bytes. Is it possible to determine the network's bottleneck with this data? If so, how?

user503842
  • 101
  • 2
  • Have you ruled out everything after your network, including the FTP server itself as the bottleneck? – Sven Aug 04 '18 at 06:58
  • how could the FTP server be the bottleneck? - can you explain this to me? .. Since the FTP client is uploading a file, the server only sends acknowledgements, right? – user503842 Aug 04 '18 at 07:09
  • Obviously, a data transfer can be only as fast as the remote site can handle, regardless of the capabilities of your side. There are countless things that might make the remote site slow (e.g. the route to it might have slow links, the server itself might have a slow or overloaded link, rate limiting, I/O overload on the disk side ....) – Sven Aug 04 '18 at 07:19
  • hmm.. ok. Well the link that connects the final router with the server is an Ethernet connection with 100 Mbit/s. Actually all existing connections have 100 Mbit/s. There is only one connection with 50 Mbit/s. And of course Wi-fi (802.11g) has a bitrate of 54 Mbit/s. I believe the Access Point is the cause of trouble here: It is the only device with a queue-size limit, there are two other routers present in the network, but they are considered ideal. It must be somehow that the FTP packets are waiting too long in the AP's queue. Does this make sense to you? – user503842 Aug 04 '18 at 07:43
  • Does the AP allow wireless A/B? Or does it only allow G/N? – Greg Askew Aug 04 '18 at 12:41

1 Answers1

0

Do not assume that components are performing ideally. Test.

Do a pure network throughput test over wireless to the FTP server, such as iperf. This removes the variable about how fast the storage is.

Temporarily plug a client, like a laptop, directly into the wired Ethernet that the AP is using. Do the FTP test and the iperf test.

  • If wired FTP is much better, the bottleneck was wireless.
  • If iperf is much better but FTP isn't, there is still a bottleneck in the FTP service.
  • If neither improves much, there is a bottleneck somewhere else on the network. Look at each hop, especially slower links like WAN/Internet.

802.11 is shared spectrum, horribly noisy, and prone to interference. You will not get close to the theoretical maximum.

John Mahowald
  • 30,009
  • 1
  • 17
  • 32