Community discussions

MikroTik App
 
bearmeister
just joined
Topic Author
Posts: 17
Joined: Tue Apr 09, 2013 6:47 am

ROS6.6 goes to 100% CPU doing 1 stream TCP test at 1Kbps ?!?

Fri Dec 06, 2013 4:58 am

Hi All,

I have an RB750GL and I am doing a bandwidth test to a directly connected RB750GL on ether1.

It seems regardless of what TCP settings I use, the CPU immediately on the "send" device goes to 100% from a baseline of 1%

This occurs regardless of how many stream or bit rate is used, for testing this CPU issue, I am using just 1 TCP stream at a pitiful 1Kbps, and it still go to straight to 100% CPU

Screenshots attached

Any feedback appreciated

Kind regards,
You do not have the required permissions to view the files attached to this post.
 
User avatar
janisk
MikroTik Support
MikroTik Support
Posts: 6263
Joined: Tue Feb 14, 2006 9:46 am
Location: Riga, Latvia

Re: ROS6.6 goes to 100% CPU doing 1 stream TCP test at 1Kbps

Fri Dec 06, 2013 4:09 pm

this is how TCP bandwidth-test works.

to test network use endpoints that are not your network infrastructure, just attached nodes to the path you want to test.
 
bearmeister
just joined
Topic Author
Posts: 17
Joined: Tue Apr 09, 2013 6:47 am

Re: ROS6.6 goes to 100% CPU doing 1 stream TCP test at 1Kbps

Tue Dec 10, 2013 11:25 am

this is how TCP bandwidth-test works.

to test network use endpoints that are not your network infrastructure, just attached nodes to the path you want to test.
Err. No. It doesn't. I'm quite familiar with a variety of test equipment from industry recognised iperf at the low cost end through to spirent carrier class testing equipment.. Suggesting that the CPU of a test chassis should go to 100% when performing a 1 kbps test, well I'm just gobsmacked and i don't even know where to go with that.

If I'm testing to the limits of the test devices then sure the CPU is obviously going to top out, but when I am sending a piddly amount of data in an attempt to test at a particular rate, you're telling me it's still okay for the CPU to sit at 100% for 1Kbps rate ? What pray tell does the device do that requires it to use 100% of it's CPU when transmitting 1Kbps of TCP traffic?

More importantly, how the heck am I supposed to do a TCP test at any rate then without CPU starving the test devices for the duration of the test?
 
onnoossendrijver
Member
Member
Posts: 486
Joined: Mon Jul 14, 2008 11:10 am
Location: The Netherlands

Re: ROS6.6 goes to 100% CPU doing 1 stream TCP test at 1Kbps

Tue Dec 10, 2013 11:53 am

Do not run a bandwidth test on a non-x86 routerboard. You should use 2 PC's connected on to the RB750 and run an iperf test between those, via the RB750.
 
InoX
Forum Guru
Forum Guru
Posts: 1966
Joined: Tue Jan 09, 2007 6:44 pm

Re: ROS6.6 goes to 100% CPU doing 1 stream TCP test at 1Kbps

Tue Dec 10, 2013 12:27 pm

receiving 1kbps - cpu 10%
sending 1kbps - cpu 100%
sending 100mbps - cpu 100%
So what is your concern? Does it hurts that cpu is starving to death? Are you having nightmares? Trouble sleeping?
Just make the test and go to sleep. Router will survive.
 
User avatar
janisk
MikroTik Support
MikroTik Support
Posts: 6263
Joined: Tue Feb 14, 2006 9:46 am
Location: Riga, Latvia

Re: ROS6.6 goes to 100% CPU doing 1 stream TCP test at 1Kbps

Tue Dec 10, 2013 12:49 pm

Err. No. It doesn't. I'm quite familiar with a variety of test equipment from industry recognised iperf at the low cost end through to spirent carrier class testing equipment.. Suggesting that the CPU of a test chassis should go to 100% when performing a 1 kbps test, well I'm just gobsmacked and i don't even know where to go with that.

If I'm testing to the limits of the test devices then sure the CPU is obviously going to top out, but when I am sending a piddly amount of data in an attempt to test at a particular rate, you're telling me it's still okay for the CPU to sit at 100% for 1Kbps rate ? What pray tell does the device do that requires it to use 100% of it's CPU when transmitting 1Kbps of TCP traffic?

More importantly, how the heck am I supposed to do a TCP test at any rate then without CPU starving the test devices for the duration of the test?
this is neither of things you named you know - it is bandwidth-test it works this way. want to load CPU run TCP bandwidth-test to itself. Only thing that can be reliably done on the testing router. In all other cases, move this load off the router you are testing.

Who is online

Users browsing this forum: Ahrefs [Bot], baragoon, xrlls and 106 guests