Recent Changes - Search:



My journals will take the place of a blog. They have moved to

Sites I take responsibility for






Places I frequent



Items for sale:


edit SideBar

Network Management and Monitoring

Include our styles below Infobox - invoke as >>infobox<< ... >><<

Codebox: - invoke as >>codebox<< ... >><<

warnbox: - invoke as >>codebox<< ... >><<

editingbox: - invoke as >>codebox<< ... >><<

noticebox: - invoke as >>codebox<< ... >><<

Page bread crumbs:

Pages by tags: (:listtags:)
Subscribe to this wiki: RSS Feed RSS or subscribe to this page for changes: RSS Feed RSS
496 articles have been published so far. Recent changes
(:addThis btn="custom":)

Speed and bandwidth testing:


For bandwidth locally or point-to-point testing I recommend Bing. I am using Bing version 1.1.3. Bing determines bandwidth on a point-to-point link by sending ICMP ECHO_REQUEST packets and measuring their round-trip times for different packet sizes on each end of the link.

host1 is supposed to be the nearest end of the link, while host2 is the other end.

I used the following command to test it on my Linux server ( to my router in the same location

(:code header=Output from make install lang=Bash wrap=80:)

 # bing -c 1 -i 2  

(:codeend:) Bing runs on Linux and Windows (I used Cygwin but binaries exists around the net)

ping and traceroute commands (standard in all unix and Microsoft Windows) should be sufficient for latency testing. If you need more than millisecond resolution though you will have to ditch windows and use a real OS. eg, here's what my Linux server manages:

(:code header=Output from make install lang=Bash wrap=80:)

 # ping -c 5
 PING ipcop ( 56(84) bytes of data.
 64 bytes from ipcop ( icmp_seq=1 ttl=64 time=0.303 ms
 64 bytes from ipcop ( icmp_seq=2 ttl=64 time=0.276 ms
 64 bytes from ipcop ( icmp_seq=3 ttl=64 time=0.277 ms
 64 bytes from ipcop ( icmp_seq=4 ttl=64 time=0.332 ms
 64 bytes from ipcop ( icmp_seq=5 ttl=64 time=0.325 ms

 --- ipcop ping statistics ---
 5 packets transmitted, 5 received, 0% packet loss, time 4004ms
 rtt min/avg/max/mdev = 0.276/0.302/0.332/0.030 ms


For faults, you can get a bit of mileage out of a PC running as a layer 2 bridge. For example here at we have a Linux box running ntop which regularly picks up stupidity on the network.

If you want anything more, buy dedicated hardware testing gear. Expect to have a large dent in your budget as a result.

Other bandwidth testing tools:

Symmetric links vs asymmetric links:

What are asymmetric links?

Taken from

"Many links between servers, and the early modems, used symmetric links. With a symmetric link the speed of data moving to your system is the same as the speed of data moving away from your system. As new high speed modems were developed to service the growing demand for the Internet links became asymmetric. For example, on a 56kilo-Baud modem, data downloads to you at 56kBaud. But your upload speed is restricted to around 28 to 33kBaud. This is because it is assume you will always download far more than you upload. The system therefore devotes more of the bandwidth of the link to download than upload. But with the development of more powerful computing, and peer-to-peer links, asymmetric connections are highly restrictive. The new broadband systems, such as ADSL, are also asymmetric. This has implications for how people can use the new broadband systems. Effectively the design of the systems assumes that everyone is a data consumer rather than a data producer or sharer. For those who wish to have a symmetric connection the readily available option is ISDN, or a faster leased line. But these connections are assumed to be primarily a tool for business, and so are far more expansive than other broadband links." (

Attach:Detecting_and_Measuring_Asymmetric_Links_in_an_IP_Network.pdf (Original link)

"Various bandwidth estimation tools have been developed. Pathchar and Bing measure per hop link bandwidth by analyzing a packet's Round Trip Time (RTT) linearity with respect to its size, at a given hop. Bprobe and Tcpanaly measures bottleneck link bandwidth based on packet-pair technique. These tools assume that the links are symmetric along the network path." - Taken from "Detecting and Measuring Asymmetric Links in an IP Network" by Wenyu Jiang, Columbia University, Computer Science Dept., Technical Report: CUCS-009-99 (above attachment)

In a nutshell, Bing does not work so great or accurately once you leave the LAN using consumer grade networks like DSL or cable modems which have asymmetric tendencies behind the scenes. The bandwidth you appear to have right now might not be what you have available (point-to-point) in ten minutes.

More tools:

Using Netperf:

Requires a client host and a server host.

The command to run the server is:

(:code header=Output from make install lang=Bash wrap=80:)

 # netserver -D -d -f -p 12865 -4


This will prevents spawns and running detached which I prefer to see what is going on. Increase the verbosity level with "-v".

Typical command to run the client:

(:code header=Output from make install lang=Bash wrap=80:)

 # netperf -H -F /var/sadm/pkg/SUNWacroread/save/121104-11/undo.Z -t TCP_SENDFILE -- -s 128K -S 128K



sinfo-Advanced Network Monitoring

Other tools:

A Small Test Tool for TCP Servers: Echoping

Bandwidth Monitor NG (Next Generation)

Create a Textual and Graphical Readout of Current Bandwidth

bmon Bandwidth Monitor

Monitor For IP Traffic without SNMP


Kevin's Public Wiki maintained and created by Kevin P. Inscoe is licensed under a
Creative Commons Attribution 3.0 United States License.

Back to my web site -

Edit - History - Print - Recent Changes - Search
Page last modified on April 22, 2013, at 07:40 PM EST