Informal network test

What data center and where were you testing to?

@eld101:

What data center and where were you testing to?

From a Linode 512 in London to another Linode 512 also in London.

I wasn't testing the internet speed, just how fast cakers fancy new switches are willing to go. It looks like we hit some kind of wall before we get to the 250 Mbits/sec bandwidth cap but we get pretty close. The CPU wasn't maxed out on either of my Linodes at any time during this test.

I don't have graphs, but I maxed out at about 236 Mbps between two Atlanta nodes.

Edit:

^Cmnordhoff@fairy:~$ date && iperf -c atlanta-130-3.linode.rdns -V
Sun Mar 10 05:46:55 UTC 2013
------------------------------------------------------------
Client connecting to atlanta-130-3.linode.rdns, TCP port 5001
TCP window size: 19.8 KByte (default)
------------------------------------------------------------
[  3] local 2600:3c02::f03c:91ff:fe70:fd08 port 40405 connected with 2600:3c02::13:3 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec   281 MBytes   236 Mbits/sec

^Cmnordhoff@foxtrot:~$ date && iperf -c f -V
Sun Mar 10 05:47:08 UTC 2013
------------------------------------------------------------
Client connecting to f, TCP port 5001
TCP window size: 20.8 KByte (default)
------------------------------------------------------------
[  3] local 2600:3c02::13:3 port 58507 connected with 2600:3c02::13:2 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec   283 MBytes   237 Mbits/sec

No graphs? But I thought we all agreed, we'd ALWAYS have graphs.

:(

I think we actually need to put the numbers together in a graph on how many did/did not agree to having graphs.

I vote numbers and graphs. I'm wondering if the graphs are totally accurate for short duration tests though.

Well, they're a five-minute average. So if your "short duration" is 5 minutes…yes? :D

I tested with iperf using TCP over IPv4 from a London Linode to an amazon machine in Ireland got 149 Mbits/sec.

That's pretty good for something that goes over the Internet.

The test was too short to generate a graph worth looking at.

@mnordhoff:

Well, they're a five-minute average. So if your "short duration" is 5 minutes…yes? :D

The short duration would have to be 10 minutes to ensure it covers a 5 minute duration. I don't think I tested for 10 minutes.

From Linode UK (512 node) to a gigabit connection in Norway.

–----------------------------------------------------------

Client connecting to xxxx.xxxxx.xxx, TCP port 5001

TCP window size: 22.5 KByte (default)

[ 3] local 178.79.166.63 port 42274 connected with xxx.xx.xxx.xxx port 5001

[ ID] Interval Transfer Bandwidth

[ 3] 0.0-40.0 sec 1.11 GBytes 239 Mbits/sec

About 15ms of network in between.

IP, TCP and ethernet overheads works out to like 4-5% for IPv4, 6-7% for IPv6 when using a 1500 byte MTU (which is used on most of the Internet) - so 237-240Mbit/s is pretty much max of what can be expected for TCP on a 250Mbit/s cap.

edit: ping round trip is 30ms - making it about 15ms of network rather.

And in the opposite direction:

–----------------------------------------------------------

Client connecting to 178.79.166.63, TCP port 5001

TCP window size: 22.5 KByte (default)

[ 3] local xxx.xx.xxx.xxx port 48575 connected with 178.79.166.63 port 5001

[ ID] Interval Transfer Bandwidth

[ 3] 0.0-60.0 sec 3.63 GBytes 520 Mbits/sec

Lets just say restoring from our remote backups is quite fast

:)

@sednet:

I tested with iperf using TCP over IPv4 from a London Linode to an amazon machine in Ireland got 149 Mbits/sec.

That's pretty good for something that goes over the Internet.

The test was too short to generate a graph worth looking at.

Perhaps the delay was caused by having to cross the English Channel. :-)

@jebblue:

@sednet:

I tested with iperf using TCP over IPv4 from a London Linode to an amazon machine in Ireland got 149 Mbits/sec.

That's pretty good for something that goes over the Internet.

The test was too short to generate a graph worth looking at.

Perhaps the delay was caused by having to cross the English Channel. :-)

The English channel is between England and France, not England and Ireland. I'm not sure what way the data goes though. Traceroute shows the traffic going via Telianet who appear to be in Sweden.

@sednet:

@jebblue:

@sednet:

I tested with iperf using TCP over IPv4 from a London Linode to an amazon machine in Ireland got 149 Mbits/sec.

That's pretty good for something that goes over the Internet.

The test was too short to generate a graph worth looking at.

Perhaps the delay was caused by having to cross the English Channel. :-)

The English channel is between England and France, not England and Ireland. I'm not sure what way the data goes though. Traceroute shows the traffic going via Telianet who appear to be in Sweden.

Yup, it was humor. :)