Netperf 返回零吞吐量以及每秒零 Trans.Rate

Question

I am using Netperf tool to benchmark throughput and latency between two VM which have private IP 10.0.1.3 and 10.0.1.13. VM 10.0.1.13

eth0      Link encap:Ethernet  HWaddr 02:3d:b2:f5:33:95
          inet addr:10.0.1.13  Bcast:10.0.1.255  Mask:255.255.255.0
          inet6 addr: fe80::3d:b2ff:fef5:3395/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1450  Metric:1
          RX packets:514888981 errors:0 dropped:0 overruns:0 frame:0
          TX packets:654103302 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:1164689163796 (1.1 TB)  TX bytes:1060437004684 (1.0 TB)

VM 10.0.1.3

eth0      Link encap:Ethernet  HWaddr 02:91:51:fa:03:08
          inet addr:10.0.1.3  Bcast:10.0.1.255  Mask:255.255.255.0
          inet6 addr: fe80::91:51ff:fefa:308/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1450  Metric:1
          RX packets:2111769 errors:0 dropped:0 overruns:0 frame:0
          TX packets:1936716 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:44230395762 (44.2 GB)  TX bytes:663995366 (663.9 MB)

I make VM 10.0.1.3 become server listen in port 12001 using the following command:

netserver -d -L 10.0.1.3 -p 12001 -4

After that, I run Netperf test from VM 10.0.1.13 with the following command:

netperf -d -H 10.0.1.3 -p 12001 -t TCP_STREAM

The connection and test running as expected but I always get value of throughput is 0:

calculate_confidence: itr  1; time 10.000263; res  0.000000
                               lcpu -1.000000; rcpu -1.000000
                               lsdm -1.000000; rsdm -1.000000
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

 87380  16384  16384    10.00       0.00

Similar to the TCP_RR test, I also received the 0 value in the Trans.Rate per sec.

calculate_confidence: itr  1; time 10.000262; res  0.000000
                               lcpu -1.000000; rcpu -1.000000
                               lsdm -1.000000; rsdm -1.000000
Local /Remote
Socket Size   Request  Resp.   Elapsed  Trans.
Send   Recv   Size     Size    Time     Rate
bytes  Bytes  bytes    bytes   secs.    per sec

16384  87380  1        1       10.00       0.00
16384  87380

I see a similar topic that related to this question, but I can not find the answer in this topic: Netperf reporting zero throughput

So, does anyone has some idea about this result.

Thanks in advance for reading my question, I hope I can find something to debug.

Answer 1

A.netperf test has two "connections." The first is the "control connection" over which information about the test setup and result is exchanged. For the benchmarking itself a "data connection" is used. The control connection will use the control port you've specified with the global "-p" option. The data connection will by default use a port number chosen by the.networking stack where the.netserver runs.

Both have to be open through firewalls for a test to be successful.

If only the control port is open, you will see the test banners get displayed because the control connection is established. Since the data connection cannot be established, that will report zero.

You can specify an explicit port number for the data connection with a test-specific "-P" option. So, if you opened a second port number, 12002, you would start the.netserver as before, and then your.netperf command would become:

netperf -d -H 10.0.1.3 -p 12001 -t TCP_STREAM -- -P,12002

That comma is important. The test-specific -P option allows specifying both the local and remote port numbers for the data connection. The remote port number follows a comma.