0

I'm looking for a way to test the bandwidth throughput of a c3.8xlarge instance.

I have two c3.8xlarge instances set up, so I imagine I could use the other to simulate the traffic.

What's the easiest way to go about it?

niemion
  • 161
  • 1
  • 8
  • iperf is simple enough, and very useful. keep in mind that you will probably not get correct values if you test between two aws instances – Dusan Bajic Jan 28 '14 at 17:39
  • @dusan.bajic Why would he not get accurate data testing between two EC2 instances? – EEAA Jan 28 '14 at 17:49
  • 1
    @EEAA, It may be because the numbers reported by iperf when testing between EC2 instances, which could even be located in the same region and zone, would be unrealistic high when compared to testing the c3.8xlarge instance with servers outside the EC2 network. – niemion Jan 28 '14 at 18:25
  • I think we need more details. Are you looking for the throughput from the instance to an outside source or strictly between two instances? – Mike Naylor Jan 28 '14 at 18:43
  • As a reference I would like to first know the throughput between the two c3.8xlarge instances. Then I want to test from an outside source. I have just performed the iperf test between the two instances, and it gives me readings at 1.72 Gbits/sec to 1.73 Gbits/sec Gbps. Why so low? Both instances have 10 GbE interfaces so I would expect much closer to 8-9 Gbps. – niemion Jan 28 '14 at 18:48
  • 1
    Are the two instances in a VPC? – EEAA Jan 28 '14 at 18:56
  • @EEAA, I haven't set up VPC, so I believe not. – niemion Jan 28 '14 at 19:00
  • Are they in the same region and availability zone? – Nathan C Jan 28 '14 at 19:03
  • If you need full bandwidth between nodes, you'll need to use [placement groups](http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/placement-groups.html). – EEAA Jan 28 '14 at 19:12
  • @Nathan C - Yes, they are in the same region and availability zone. I will have to read about placement groups, do you think it is needed to see speeds over 1.7 Gbps? – niemion Jan 28 '14 at 19:59
  • /etc/sysctl.conf/limits.d/90-nproc.conf (which overrides values in /etc/security/limits.conf) has "* soft nproc 1024". Should setting this number higher or even to unlimited do anything with regards to eliminating this 1.73 Gbps limit? – niemion Jan 29 '14 at 00:17
  • If you test from another EC2 instance, you'll only be testing the in-datacenter bandwidth. Even if your 2nd server is in a different EC2 regions - you will be testing over Amazon-optimized infrastructure. So if you want to see how much bandwidth you have for serving end users across the internet, you need to be sure your test measures traffic in the same way. – CMerrill Jan 29 '14 at 14:05
  • Thank you @CMerrill, first I'm trying to find out if there are limitations on the instance that makes it impossible to archive speeds over 1.73 Gbits/sec. When I find out what to change in order to take full advantage of the 10 GbE adapter on the c3.8xlarge instance. Then when I figure that out I will test from outside the Amazon environment. – niemion Jan 29 '14 at 16:11
  • More about the 1.73 Gbps cap here: http://serverfault.com/questions/570879/1-73-gbps-at-best-on-an-amazon-ec2-10-gigabit-instance – niemion Feb 01 '14 at 13:25

1 Answers1

1

few options..

  • (nu)ttcp
  • netcat and dd
  • iperf as someone already mentioned

ttcp is pretty simple. on one machine do nuttcp -S and on the other do nuttcp -v -v -i10 otherhostname

-i is the interval to print results

skrewler
  • 308
  • 1
  • 3
  • 13