06-14-2013 04:12 AM
Hi folks,
I need to test failover scenario, can anybody suggest a software that i can send continuous burst and soft provide the report of packet drops and switch over delay.
You input will be appreciated
Regards
Solved! Go to Solution.
06-14-2013 05:37 AM
Hi,
It depends what platforms you have access to, but with Linux/UNIX I tend to use nuttcp these days. I'm currently running the version 7.2.1 from the beta directory. There are Windows binaries built as well if you don't have Linux, but these are several versions behind.
This program is a client/server based application that can be used to generate TCP or UDP traffic at specific rates. If you're looking to determine the level of packet loss, then the UDP testing would be the most appropriate as this will tell you how many packets have been dropped.
The following is an example of sending UDP traffic (-u) at a rate of 50Mbps (-R50M) with packet size of 384-bytes (-l384) between two hosts called rhel4 and rhel5.
[sfuller@rhel8 ~]$ nuttcp -u -i1 -R50M -l384 rhel4 rhel5
5.9692 MB / 1.00 sec = 50.0100 Mbps 0 / 16300 ~drop/pkt 0.00 ~%loss
5.9594 MB / 1.00 sec = 50.0120 Mbps 0 / 16273 ~drop/pkt 0.00 ~%loss
5.9594 MB / 1.00 sec = 50.0123 Mbps 0 / 16273 ~drop/pkt 0.00 ~%loss
5.9652 MB / 1.00 sec = 50.0117 Mbps 0 / 16289 ~drop/pkt 0.00 ~%loss
5.9652 MB / 1.00 sec = 50.0113 Mbps 0 / 16289 ~drop/pkt 0.00 ~%loss
5.9594 MB / 1.00 sec = 50.0123 Mbps 0 / 16273 ~drop/pkt 0.00 ~%loss
5.9652 MB / 1.00 sec = 50.0111 Mbps 0 / 16289 ~drop/pkt 0.00 ~%loss
5.9590 MB / 1.00 sec = 50.0093 Mbps 0 / 16272 ~drop/pkt 0.00 ~%loss
5.9594 MB / 1.00 sec = 50.0125 Mbps 0 / 16273 ~drop/pkt 0.00 ~%loss
59.6060 MB / 10.00 sec = 50.0113 Mbps 99 %TX 7 %RX 0 / 162764 drop/pkt 0.00 %loss
If you go to Phil Dykstra's nuttcp page, there's a quick start guide and also a PDF that has a lot more detail.
Regards
06-14-2013 05:37 AM
Hi,
It depends what platforms you have access to, but with Linux/UNIX I tend to use nuttcp these days. I'm currently running the version 7.2.1 from the beta directory. There are Windows binaries built as well if you don't have Linux, but these are several versions behind.
This program is a client/server based application that can be used to generate TCP or UDP traffic at specific rates. If you're looking to determine the level of packet loss, then the UDP testing would be the most appropriate as this will tell you how many packets have been dropped.
The following is an example of sending UDP traffic (-u) at a rate of 50Mbps (-R50M) with packet size of 384-bytes (-l384) between two hosts called rhel4 and rhel5.
[sfuller@rhel8 ~]$ nuttcp -u -i1 -R50M -l384 rhel4 rhel5
5.9692 MB / 1.00 sec = 50.0100 Mbps 0 / 16300 ~drop/pkt 0.00 ~%loss
5.9594 MB / 1.00 sec = 50.0120 Mbps 0 / 16273 ~drop/pkt 0.00 ~%loss
5.9594 MB / 1.00 sec = 50.0123 Mbps 0 / 16273 ~drop/pkt 0.00 ~%loss
5.9652 MB / 1.00 sec = 50.0117 Mbps 0 / 16289 ~drop/pkt 0.00 ~%loss
5.9652 MB / 1.00 sec = 50.0113 Mbps 0 / 16289 ~drop/pkt 0.00 ~%loss
5.9594 MB / 1.00 sec = 50.0123 Mbps 0 / 16273 ~drop/pkt 0.00 ~%loss
5.9652 MB / 1.00 sec = 50.0111 Mbps 0 / 16289 ~drop/pkt 0.00 ~%loss
5.9590 MB / 1.00 sec = 50.0093 Mbps 0 / 16272 ~drop/pkt 0.00 ~%loss
5.9594 MB / 1.00 sec = 50.0125 Mbps 0 / 16273 ~drop/pkt 0.00 ~%loss
59.6060 MB / 10.00 sec = 50.0113 Mbps 99 %TX 7 %RX 0 / 162764 drop/pkt 0.00 %loss
If you go to Phil Dykstra's nuttcp page, there's a quick start guide and also a PDF that has a lot more detail.
Regards
06-14-2013 08:25 AM
06-14-2013 09:20 AM
I agree with mfurnival that iperf is also a good tool for traffic generation, and in fact was my tool of choice before I turned to nuttcp.
I prefer nuttcp now as with it running from xinetd I can fire up a stream between different hosts without visiting those hosts. This is what you can see in the example above where I'm on a host called rhel8 and start a stream between the hosts rhel4 and rhel5.
The other thing that nuttcp can do for me that iperf can't is that I can set a TCP stream to transmit as a fixed rate in the same way we can for the UDP streams. I don't get this option with iperf; it will always go as fast as TCP will allow it over the network path being used.
The one area where iperf is still the winner in my mind is for multicast. While multicast support is available in nuttcp with the -g option, you have to use multicast groups in the range 231.x.y.z, which is no use if you working in an environment where some other multicast groups ranges are being used.
As with any toolbox, it's good to have a variety of tools in there, and know which one to use and when.
Regards
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide