ndt-users list created

[Peter Van Epp <>]

On Fri, Aug 25, 2006 at 09:11:37AM -0400, Richard Carlson wrote:
> Hi Peter;
> 
> Us the "More Details" button and let me know what the variables
>         RcvWinScale
>         SndWinScale
>         WinScaleRcvd
>         WinScaleSent
> 
> report?
> 
> Rich
> 

        Well this morning all works correctly (although I just checked the 
IPs in the trace and we were going to the new server when it reported no 
window scaling last night). It may be a browser issue as I find I need to 
kill and restart firefox in order to switch between ndt servers or else odd
things occur. It appears that the new version is more correct than v5.3.3e.
On the Mac with default window sizes v5.3.3e reports no window scaling 
despite the fact that it is enabled but just not in use (presumably because
of 32K buffers not needing it). v5.4.8 reports window scaling on for both
servers:


Mac large buffers enabled new ndt  (PowerPC platform gig link, 100 megs on 
                                    client)

test4:/Users/vanepp root# sysctl -w kern.ipc.maxsockbuf=8000000
kern.ipc.maxsockbuf: 262144 -> 8000000
test4:/Users/vanepp root# sysctl -w net.inet.tcp.sendspace=4000000
net.inet.tcp.sendspace: 32768 -> 4000000
test4:/Users/vanepp root# sysctl -w net.inet.tcp.recvspace=4000000
net.inet.tcp.recvspace: 32768 -> 4000000

        This appears to really screw up throughput though, but it does 
enable window scaling on both servers :-)

TCP/Web100 Network Diagnostic Tool v5.4.8
click START to begin
Checking for Middleboxes . . . . . . . . . . . . . . . . . .  Done
checking for firewalls . . . . . . . . . . . . . . . . . . .  Done
running 10s outbound test (client-to-server [C2S]) . . . . . 80.88Mb/s
running 10s inbound test (server-to-client [S2C]) . . . . . . 28.60Mb/s
The slowest link in the end-to-end path is a 100 Mbps Full duplex Fast 
Ethernet subnet
[S2C]: Excessive packet queuing detected

click START to re-test

WEB100 Enabled Statistics:
Checking for Middleboxes . . . . . . . . . . . . . . . . . .  Done
checking for firewalls . . . . . . . . . . . . . . . . . . .  Done
running 10s outbound test (client-to-server [C2S]) . . . . . 80.88Mb/s
running 10s inbound test (server-to-client [S2C]) . . . . . . 28.60Mb/s

        ------  Client System Details  ------
OS data: Name = Mac OS X, Architecture = ppc, Version = 10.3.9
Java data: Vendor = Apple Computer, Inc., Version = 1.4.2_09

        ------  Web100 Detailed Analysis  ------
100 Mbps FastEthernet link found.
Link set to Full Duplex mode
No network congestion discovered.
Good network cable(s) found
Normal duplex operation found.

Web100 reports the Round trip time = 72.31 msec; the Packet size = 1460 
Bytes; and 
No packet loss was observed.
C2S throughput test: Packet queuing detected: 2.94%
S2C throughput test: Excessive packet queuing detected: 42.53%

Web100 reports TCP negotiated the optional Performance Settings to: 
RFC 2018 Selective Acknowledgment: OFF
RFC 896 Nagle Algorithm: OFF
RFC 3168 Explicit Congestion Notification: OFF
RFC 1323 Time Stamping: OFF
RFC 1323 Window Scaling: ON

Server 'sniffer1.ucs.sfu.ca' is not behind a firewall. [Connection to the 
ephemeral port was successful]
Client is probably behind a firewall. [Connection to the ephemeral port 
failed]
Information: Network Middlebox is modifying MSS variable
Server IP addresses are preserved End-to-End
Client IP addresses are preserved End-to-End

WinScaleRcvd: 5
WinScaleSent: 8
RcvWinScale: 8
SndWinScale: 5


old ndt  (Intel platform, opposite endian)

WEB100 Enabled Statistics:
Checking for Middleboxes . . . . . . . . . . . . . . . . . .  Done
running 10s outbound test (client to server) . . . . . 97.00Mb/s
running 10s inbound test (server to client) . . . . . . 50.58Mb/s

        ------  Client System Details  ------
OS data: Name = Mac OS X, Architecture = ppc, Version = 10.3.9
Java data: Vendor = Apple Computer, Inc., Version = 1.4.2_09

        ------  Web100 Detailed Analysis  ------
Interprocess communications failed, unknown link type.
Link set to Half Duplex mode
Information: throughput is limited by other network traffic.
Good network cable(s) found
Normal duplex operation found.

Web100 reports the Round trip time = 60.93 msec; the Packet size = 1460 
Bytes; and 
No packet loss - but packets arrived out-of-order 29.62% of the time
This connection is receiver limited 66.81% of the time.
This connection is network limited 33.07% of the time.

Web100 reports TCP negotiated the optional Performance Settings to: 
RFC 2018 Selective Acknowledgment: OFF
RFC 896 Nagle Algorithm: ON
RFC 3168 Explicit Congestion Notification: OFF
RFC 1323 Time Stamping: OFF
RFC 1323 Window Scaling: ON
Packet size is preserved End-to-End
Server IP addresses are preserved End-to-End
Client IP addresses are preserved End-to-End

WinScaleRcvd: 6
WinScaleSent: 8
SndWinScale: 6
RcvWinScale: 8

then go back to the default buffer sizes by rebooting (it didn't like me 
setting them back lower :-)). 

test4:/Users/vanepp root# sysctl -w kern.ipc.maxsockbuf=262144
kern.ipc.maxsockbuf: 8000000 -> 262144
test4:/Users/vanepp root# sysctl -w net.inet.tcp.sendspace=32768
net.inet.tcp.sendspace: 4000000 -> 32768
test4:/Users/vanepp root# sysctl -w kern.ipc.maxsockbuf=32768 
kern.ipc.maxsockbuf: 262144 -> 32768
test4:/Users/vanepp root# 

        Then reboot because it didn't like this :-)

TCP/Web100 Network Diagnostic Tool v5.4.8
click START to begin
Checking for Middleboxes . . . . . . . . . . . . . . . . . .  Done
checking for firewalls . . . . . . . . . . . . . . . . . . .  Done
running 10s outbound test (client-to-server [C2S]) . . . . . 90.96Mb/s
running 10s inbound test (server-to-client [S2C]) . . . . . . 79.71Mb/s
The slowest link in the end-to-end path is a 100 Mbps Full duplex Fast 
Ethernet subnet

click START to re-test

WEB100 Enabled Statistics:
Checking for Middleboxes . . . . . . . . . . . . . . . . . .  Done
checking for firewalls . . . . . . . . . . . . . . . . . . .  Done
running 10s outbound test (client-to-server [C2S]) . . . . . 90.96Mb/s
running 10s inbound test (server-to-client [S2C]) . . . . . . 79.71Mb/s

        ------  Client System Details  ------
OS data: Name = Mac OS X, Architecture = ppc, Version = 10.3.9
Java data: Vendor = Apple Computer, Inc., Version = 1.4.2_09

        ------  Web100 Detailed Analysis  ------
100 Mbps FastEthernet link found.
Link set to Full Duplex mode
No network congestion discovered.
Good network cable(s) found
Normal duplex operation found.

Web100 reports the Round trip time = 73.31 msec; the Packet size = 1460 
Bytes; and 
No packet loss was observed.
C2S throughput test: Packet queuing detected: 0.64%
S2C throughput test: Packet queuing detected: 7.46%

Web100 reports TCP negotiated the optional Performance Settings to: 
RFC 2018 Selective Acknowledgment: OFF
RFC 896 Nagle Algorithm: OFF
RFC 3168 Explicit Congestion Notification: OFF
RFC 1323 Time Stamping: OFF
RFC 1323 Window Scaling: ON

Server 'sniffer1.ucs.sfu.ca' is not behind a firewall. [Connection to the 
ephemeral port was successful]
Client is probably behind a firewall. [Connection to the ephemeral port 
failed]
Information: Network Middlebox is modifying MSS variable
Server IP addresses are preserved End-to-End
Client IP addresses are preserved End-to-End


RcvWinScale: 8
SndWinScale: 5
WinScaleRcvd: 5
WinScaleSent: 8

old ndt server   both server and client on 100 meg links  Note window scaling
is reported off here despite no change from the run above to the other ndt
server where it is reported on which may be what we were seeing last night
although Lixin has 1 meg buffers set on the machine that was failing. 

TCP/Web100 Network Diagnostic Tool v5.3.3e
click START to begin
Checking for Middleboxes . . . . . . . . . . . . . . . . . .  Done
running 10s outbound test (client to server) . . . . . 88.60Mb/s
running 10s inbound test (server to client) . . . . . . 91.60Mb/s
Server unable to determine bottleneck link type.

click START to re-test

WEB100 Enabled Statistics:
Checking for Middleboxes . . . . . . . . . . . . . . . . . .  Done
running 10s outbound test (client to server) . . . . . 88.60Mb/s
running 10s inbound test (server to client) . . . . . . 91.60Mb/s

        ------  Client System Details  ------
OS data: Name = Mac OS X, Architecture = ppc, Version = 10.3.9
Java data: Vendor = Apple Computer, Inc., Version = 1.4.2_09

        ------  Web100 Detailed Analysis  ------
Interprocess communications failed, unknown link type.
Link set to Full Duplex mode
No network congestion discovered.
Good network cable(s) found
Normal duplex operation found.

Web100 reports the Round trip time = 5.41 msec; the Packet size = 1460 Bytes; 
and 
No packet loss - but packets arrived out-of-order 0.57% of the time
This connection is receiver limited 98.59% of the time.
This connection is network limited 1.3% of the time.

Web100 reports TCP negotiated the optional Performance Settings to: 
RFC 2018 Selective Acknowledgment: OFF
RFC 896 Nagle Algorithm: ON
RFC 3168 Explicit Congestion Notification: OFF
RFC 1323 Time Stamping: OFF
RFC 1323 Window Scaling: OFF
Packet size is preserved End-to-End
Server IP addresses are preserved End-to-End
Client IP addresses are preserved End-to-End

WinScaleRcvd: 0
WinScaleSent: 8
RcvWinScale: 8
SndWinScale: 0

Peter Van Epp / Operations and Technical Support 
Simon Fraser University, Burnaby, B.C. Canada