The performance of your network connection is a function
of two variables: network delay and network throughput.
Network delay
Network delay is caused by several factors inherent in modern
telecommunications systems, as described below.
Propagation delay: The time to transmit data along
the physical path between each of the switches and routers
involved in the complete communications path. It is calculated
by dividing the total physical distance of the circuits by
the speed of light (186,000 miles per second in a vacuum).
Switch delay: The time required for each router to
read and forward each IP datagram.
Packet transmit time: The time required to clock
one entire IP datagram over a serial link. This time is dependent
on the physical speed of the circuit. For example, if the
user sends a 1,000-byte packet on a 56 kbps line, it will
take (1000*8)/(56000) = 143 ms.
Queuing delay: A network state in which the number
of incoming IP datagrams
exceeds the processing capability of the receiving router.
Routers will queue
datagrams until additional routing cycles become available.
Routers with lower
processing power will suffer greater queuing delay, all else
being equal. With
TCP/IP, routers will drop datagrams held for too long in queue.
The above factors combine to define the end-to-end delay
across the network.
Network throughput
Throughput in a Wide Area Network is defined as "the
amount of data that can be transmitted from one location to
another in a fixed amount of time".
Throughput is generally measured at the application level
by measuring the time it takes to complete a file transfer
or download.
How delay and throughput interrelate
Throughput observed over a network connection is inversely
related to network
delay. If delay goes down, the effective throughput goes up.
If delay goes up,
effective throughput goes down.
Network delay is related to serial line distance and inversely
related to serial-line
speed. While serial-line speed and distance are factors that
affect the throughput you observe, the packet and window size
can easily be the dominant factor in the throughput equation,
especially if they are not set for maximum performance, as
described below.
The MTU, Maximum Transmission Unit, refers to the maximum
number of bytes in a packet that cyberMIND will send through
its network at one time. Packets that exceed the MTU are fragmented
into smaller packets. Avoiding this fragmentation is a good
idea because it introduces some delay across the network.
You can avoid fragmentation by setting your routers to an
MTU value of 1500 bytes for T1 connections and below, and
to 4470 bytes for connections above T1.
IP packets that meet these MTUs will not be fragmented in
cyberMIND’s network unless the destination’s interface
is, for example, a T1, and the originating connection is 6
Mbps, In other words, 4470-byte MTUs will be broken down to
fit into the 1500-byte MTU of the T1 connection.
Many other factors can affect your throughput, including
traffic from other ISPs and the performance of their networks.
|
