Netest output for general measurement

• Read output for default mode
• Read output for -abw mode

1. Default mode (no option) on a buggy Solaris
2. tuned case 1 according to netest advisory
5. measuring available bandidth capable to do it
6. measuring available bandidth maximum throughput only

  General statistics:

The first line shows the average bandwidth and time used in this test. The second line shows TCP throughput in minimum/average/maximum Mb/sec.

  General section:

Reports the RTT and Burst size, which is the minimum ability of continuously receiving data at a router or the destination host on the path.

  UDP section:

The first two lines report single stream UDP throughput and double stream UDP throughput. The next line -- "Use multiple streams to gain UDP throughput" is either "recommended" or "not recommended" according to if double stream UDP can improve the throughput.

  TCP section:

The first three lines are always printed. The following two lines depend on the third line of this section. If "Use parallel TCP streams:" is "not recommended", then the section ends. If it is "recommended", then, one or more lines will be printed:

N TCP streams can improve the throughput from this host
recommends to use N TCP streams on this host to improve TCP performance.

M TCP streams may maximize the throughput from this subnet
recommends to use M TCP streams on this subnet (this host + other hosts) to gain maximum TCP throughput. If this line is NOT printed, it means that this host can use N TCP stream can gain the maximum throughput over this path.

TCP Burst rate is current an experimental parameter for TCP tuning.

  Advisory section:

Network status

	able to use:	shows network connectivity.
	reliable:	shows network basic data transfer ability.
	performance:	shows network performance feature.

If "able to use" is not True, it means that you may ping a host, but you are not able to connect to it via ssh/rsh etc.

If one or more of them is not True or Healthy, the netest will print further advisory information to help users to do further diagnosis.

For example:

	Please run "pipechar -P 0 xxx.yyy.zz.hhh" to find the bottleneck
	router and run "whois node-name" to find its sys-admin contact,
	then contact them to increase the bandwidth for you if you think
	the bandwidth is not enough for you :-)

Case 1:

The output from netest client running in default mode:

Solaris% netest -t xxx.yyy.gov

local CPU speed 336.00 MHz sparcv9
usable  local mem  1024 MBytes
mem bus speed       228 MBps
max mem copy speed  228 MBps
min mem copy speed   22 MBps
NIC speed       271.345 Mb/s
CPU power       56
sub_cost         186 ns
 ts_call         437 ns
sys-call         328 ns
delay_one_us2 = 0.000001
IP = 160.91.192.165 (0xffbef0fc)

Remote server is Ver2 RC-10 debug 2003-02-28
Linux xxx.yyy.gov 2.4.20-web100
has 1 thread running at Func0
usable remote mem   503 MBytes
mem bus speed       849 MBps
max mem copy speed  590 MBps
min mem copy speed  571 MBps
NIC speed       468.233 Mb/s
CPU power       617

258048 Bytes dropped (0.1845 %) ReTx 0

===== Testing results: to xxx.yyy.gov =====
--- General statistics ---
Average bandwidth used in testing: 8.1156 Mb/s in 131.4571 sec
Maximum throughput 258.774 Mb/s 

TCP transf rate:        min 6.3491 / avg 6.7888 / max 226.3634 Mb/s

--- General section ---
        Round Trip Time (RTT):          66.4000 ms
burst size(1472) 6029312 : 4358144
        Maximum burst size:             6029312 Bytes (pkt size 1472)

--- UDP section ---
        Single stream UDP throughput:   258.7744 Mbps
        Multi- stream UDP throughput:   258.7744 Mbps
        Use of multiple streams to gain UDP throughput is not recommended

--- TCP section ---
BDP T 503294 U 18017284 RTT 0.066400
burst size(1472) 6029312 : 4358144
        Optimal TCP Window size:        2970166 Bytes
        Single stream TCP throughput:   7.2286 Mbps
            Use of parallel TCP streams:        is recommended
                65 TCP streams can fully utilize the Available bandwidth (win = 786432)

--- Advisory section ---
    Network status:
        usable  :       True
        reliable:       True
        performance:    0x20: The problem currently isn't understandable

        Kernel may need patch
Please rerun test with -FullD or -win 524288
r1NIC 468.23 (rMem 1888.0) l1NIC 271.34 (lMem 729.6) Mb/s
Peak throughput is around 271.345 Mb/s  limited by local kernel

Case 2:

The output from netest client running with advisory options:

Solaris% netest -t xxx.yyy.gov -FullD

28672 Bytes dropped (0.0147 %) ReTx 0

===== Testing results: to xxx.yyy.gov =====
--- General statistics ---
Average bandwidth used in testing: 19.8127 Mb/s in 75.3439 sec
Maximum throughput 266.796 Mb/s

TCP transf rate:        min 56.5270 / avg 110.4390 / max 239.7764 Mb/s

--- General section ---
        Round Trip Time (RTT):          66.7000 ms
        Maximum burst size:             6709248 Bytes (pkt size 28672)

--- UDP section ---
        Single stream UDP throughput:   266.7961 Mbps
        Multi- stream UDP throughput:   266.7961 Mbps
        Use of multiple streams to gain UDP throughput is not recommended

--- TCP section ---
BDP T 9511831 U 18659723 RTT 0.066700
        Optimal TCP Window size:        2820407 Bytes
        Single stream TCP throughput:   135.9998 Mbps
            Use of parallel TCP streams:        is recommended
                 3 TCP streams can fully utilize the Available bandwidth (win = 1048576)

--- Advisory section ---
    Network status:
        usable  :       True
        reliable:       True
        performance:    Good

r1NIC 468.23 (rMem 1888.0) l3NIC 279.76 (lMem 742.4) Mb/s
Peak throughput is around 266.796 Mb/s  limited by  local memory