mEaSUrInG anD mOnITOrInG network round-trip
time (RTT) is important for multiple reasons: it allows
network operators and end users to understand
their network performance and help optimize their
environment, and it helps businesses understand
the responsiveness of their services to sections of their
user base.
Measuring network RTT is also
important for Transmission Control
Protocol (TCP) stacks to help optimize
bandwidth usage. TCP stacks on end
hosts optimize for high performance
by passively measuring network RTTs
using widely deployed TCP timestamp
options carried in TCP headers. This information, if utilized, carries some distinct operational advantages for services and applications: hosts do not need
to launch out-of-band Internet Control
Message Protocol (ICMP) echo requests
(pings), nor do they need to embed timing information into application traffic.
Instead, hosts can passively measure
R TT representative of full-path network
latency experienced by TCP traffic.
Understanding network delay is
key to understanding some important
aspects of network performance. The
time taken to traverse the network between two hosts affects how responsive
services are, and it affects the effective
bandwidth available to end hosts. Measuring this information passively on
servers can help provide a fine-grained
indication of service responsiveness
from the customer’s perspective, and
it simultaneously offers a network-distance metric for customers that is
more useful than coarse-grained and
often inaccurate IP-based geolocation.
Measuring the RTT to many hosts or
customers is nontrivial. One solution
is active probing, in the form of ICMP
Passively
measuring tCP
Round-trip
times
Doi: 10.1145/2507771.2507781
Article development led by
queue.acm.org
A close look at round-trip time measurements
with the Transmission Control Protocol.
BY stePhen D. stRoWes