Quote:
Originally Posted by jh valley
i do a speed test about once a week and hover around 20 down; i'm curious as to why when i go and download an album (say it's 70mb), i only download at like 1mbps. why doesn't that 20 down apply to actual downloads? is there anything i can do to 'fix' this? thanks in advance for any insight  |
You can throw speedtest.net out the window for the most part. If you get 20Mbit/s that translates to 2.5MiB/s. Also keep in mind that if you have comcast powerboost or the equivalent, speedtest is largely going to be tricked by powerboost. Powerboost only applies to the first 10MiB or so.
http://customer.comcast.com/Pages/FA...4-051cd5feacf0
Below is a tcpdump of a session with speedtest.net.
Code:
osx ~ % uname -a
Darwin osx 9.8.0 Darwin Kernel Version 9.8.0: Wed Jul 15 16:55:01 PDT 2009; root:xnu-1228.15.4~1/RELEASE_I386 i386
osx ~ % sudo tcpdump -w tcpdump.en1 -i en1
Password:
tcpdump: listening on en1, link-type EN10MB (Ethernet), capture size 96 bytes
18048 packets captured
18052 packets received by filter
0 packets dropped by kernel
Note that no packets were dropped. Almost all the packets were sent/received from port 9091 (xmltec-xmlmail according to /etc/services) remotely to my machine port 55622/55623. The total transfer size was approximately 15.5MiB for the download part.
Code:
osx ~ % tcpdump -r tcpdump.en1 | perl -nle '$i += $1 if /xmlmail > osx\.\d+: [SP.] \d+:\d+\((\d+)\)/; END { print $i / 1024**2 }'
reading from file tcpdump.en1, link-type EN10MB (Ethernet)
15.560188293457
Total time from start to finish for the download was ~10 seconds.
Code:
02:14:20.473769 IP 69.38.192.34.xmltec-xmlmail > osx.55622: S 2591463299:2591463299(0) ack 2223330138 win 5792 <mss 1
460,sackOK,timestamp 373354748 673080267,nop,wscale 0>
to
Code:
02:14:30.725108 IP 69.38.192.34.xmltec-xmlmail > osx.55622: . 8138151:8139599(1448) ack 5275 win 16080 <nop,nop,times
tamp 373355772 673080369>
The point is most of the speedtest.net download is covered by powerboost, so you don't get an accurate rate anyway. I also don't believe 15MiB is enough data to accurately measure sustained throughput.