“My connection doesn’t seem as fast as it should
be”
There are many tools on the Internet and many test sites that
claim to be able to test the performance of your connection.
Unfortunately most of the “Test Sites” do not have the capability to
test accurately. In this document, we will show you how to test
performance accurately, and actually determine where the performance
bottleneck exists.
Things that will definitely impede performance: 1: Multiple Computers on your connection: If you have multiple
computers on a connection, then your performance will be divided by
the amount of bandwidth that the other machines are using.
2: Firewalls – Firewalls control and inspect packets that are
passing through and from your network. Each packet passes through
has to be fully inspected and analyzed by the firewall device or
software. This will generally cause a performance hit of 5% to 15%
on your connection.
3: Proxy Servers – All Proxy servers, especially servers based on
Windows NT, 95, 98 or Windows 2000 will impede performance
significantly on all non-cached items. Linux proxy servers, or
specialize proxy servers a generally faster, but you will always
take a performance hit with proxy technology. These will speed up
performance for
4: Network Address Translation – Network Address Translation will
impede performance because it causes the router handle the
communications for processing Internet Traffic, this increases the
CPU Load on the router, thus degrading performance.
5: Your Computer – DSL, T1, and other high speed connections are
capable of outpacing many computer’s capability to process the
incoming web pages, (especially complex web pages, and multimedia).
7: Your Network – Networks that are obtaining a medium to high
amount of collisions that are connected to a DSL modem or T1 line
will slow down performance.Components of Performance:
Most people do not understand what performance is comprised of…..
There is a lot of confusion.
The Internet is just like the public highway system. For instance,
you can (legally) drive 70 MPH in some places and 55 MPH in others
on the highway, this function is what we call “speed”. Another
important component is how wide the highway is, I.E. 2 lanes, 3
lanes or 4 lanes or more. Based upon this we can determine how many
cars can drive the speed (70MPH) that the highway is rated for. If
there are too may cars on the highway, then there will not be enough
room for them to drive fast, and the cars will have to slow down, we
call this feature “congestion”. Also at many points you need to get
off of the highway, or go through an interchange, depending on how
many cars are going through this, and how the interchange was built
will determine how quickly you can switch to a new highway, or get
off the exit ramp, this feature is called “latency”. On the Internet
Latency refers to the routers and switches that put traffic onto
other on or off ramps. Another component of performance is related
to “duplexing”, which using the highway analogy, refers the ability
of the road to take two way or one way traffic. If traffic can only
go in one direction at any particular moment, it is called
“half-duplex” and if cars can go both ways on the road at the same
time, this is called “full duplex” Together these three
components compromise the “Performance” of the Internet. It is
important to understand that these components have to be
synchronized.
Here is a clear definition of the performance characteristics:
SPEED: How fast is the connection (i.e., 128Kb, 256kb).
CONGESTION: How much of the connection, or connections are already
in use.LATENCY: How much delay exist where the switches, routers, and
modems process traffic.
DUPLEXING: The bandwidth available is it full-duplex of half duplex
Comparison of the “SPEED” of different technologies:
|
TYPE OF
CONNECTION |
THEORETICAL
SPEED (Kilobit) |
REALISTIC DOWNLOAD SPEED (Kilobyte),
ASSUMING NO OTHER TRAFFIC OR CONGESTION, AND NO
UPLOADING |
NOTES |
| Analog Modem |
33.6 |
3 k/sec |
|
| v.90 Analog Modem |
53.2 |
4.5 k/sec |
Most modems can connect at
42-48bps |
| ISDN single Channel |
64 |
5.8 k/sec |
|
| ISDN dual Channel |
128 |
11.5 k/sec |
Limited by Multi-Link PPP |
DSL 768/768
(half duplex) |
768 |
69 k/sec |
|
DSL 960/960
(half duplex) |
960 |
87 k/sec |
|
| DSL 1632/1632 (half duplex) |
1632 |
148 k/sec |
|
Full T-1 Line
(full duplex) |
1540 |
135 k/sec |
Usually limited by router
performance and T1 line overhead. Usually cannot be
obtained by a single stream of data. |
10 Mb Ethernet
(if full duplex) |
100000 |
9500k/sec |
Impossible to obtain with
one computer |
| OC-3 |
155000 |
15000 k/sec |
Backbone |
| Gigabit Ethernet |
1000000 |
90000 k/sec |
Backbone |
| OC-48 |
2480000 |
240000 k/sec |
Backbone |
| OC-192 |
99200000 |
9800000 k/sec |
Backbone |
The difference in KB and Kb:
When people are looking at speed one of the common problems is
that they do not understand the difference in Kb (K with a
little b) and KB (K with a big B). Kb is Kilobit, and KB means
Kilobyte. A kilobit is eight or nine times smaller than a
kilobyte.
TOOLS:
Ping Utility: This will give you an idea
if there is a performance issue between you and a particular
site. Generally a good ping time for off site servers is 150
Milliseconds or less.
Note: Some sites do not allow you to ping
them due to firewalling on their network..
Example below tests your connection
TraceRoute Utility: This will tell you
the relay time between different numbered routers or switches
that are used on the Internet. You can run a physical trace from
your PC directly to another IP address on the Internet.
Generally having every connection at 150Ms or less is extremely
good. When you see slow speeds on certain links it usually
suggests where the problem is.
|