Specifications

CHAPTER 2. NETWORK PERFORMANCE MEASUREMENTS SYSTEMS 6
accuracy of the measurements is good (around 10us) but depends on the operating system.
The traffic generated is for 10/100Mbps links.
2.2 Synchronization for long distance measurements
Time synchronization is a critical piece of infrastructure for any distributed system. In the
case of network research, we need synchronized clocks in order to accurately measure delays
in a distributed network that may span over a wide geographical area.
Our goal is to do one-way network measurements on a network with nodes which are
geographically far apart. This requires synchronized clocks on all the nodes involved in
the measurements. The maximum difference between two clocks should be less than 1
microsecond. This precision is needed for running tests over Gigabit Ethernet links.
The delay between end nodes is measured in the following way: each packet generated by
a node is marked with a time stamp (value of a clock counter) by the sending node and
with another time stamp when it arrives at the destination node. In order to measure the
delay caused by the network we must make the difference between these two time stamps.
For this result to be meaningful we must have the clocks in the two end nodes perfectly
synchronized (Figure 2.2).
time = t2
one way delay = t2 − t1
time = t1
Figure 2.2: Measuring one way delay between two end nodes
In the following we shall present the clock synchronization methods that are available on the
market today. Expected performance is discussed. Most of the methods involve network
time transfer protocols that provide accuracies of the order of 1-10 milliseconds. When
used with GPS and direct connection time transfer techniques they can provide accuracies
of the order of 1-10 microseconds.
2.2.1 The Network Time Protocol
The most common method of synchronizing computer clocks in a network is based on the
Network Time Protocol. This protocol is designed to distribute accurate and reliable time
information to systems operating in diverse and widely distributed inter-networked envi-
ronments. The system is based on a network of time servers operating in a self-organizing,