A major problem with file transmission between two or more
points results from latency, which comes from various sources,
including the hardware (switches, routers) and the media that you
are using. Latency is also increased by the distance a packet needs
to travel from point A to point B. This white paper highlights the
shortcomings of TCP protocol for WAN transmission purposes and
looks at your options and what you can do to address those
challenges.
Introduction
Long gone are the days where work ends at 5:00pm, or that, in
order to access a client file while on the road, you had to pull
out a physical file folder from your briefcase.
The name of the game these days is mobility and accessibility.
CIOs and other decision makers want to achieve worker productivity
by giving employees the tools they need anywhere at any time.
Unfortunately, those who make these promises sometimes don't ask
for your opinion first, and you are now tasked to make it
happen.
Software developers write code to power applications that drive
today's businesses. Whether you are looking at ubiquitous
applications such as Microsoft Word or your own in-house software,
chances are that they are not very WAN-friendly. There may be
different reasons for this, but most programmers do not take
networks into consideration when writing code, which ends up
creating very "chatty" applications.
That being said, the focus of this paper is to highlight the TCP
protocol's shortcomings for WAN transmission purposes. Let's look
at your options and what you can do to address those
challenges.
Understanding Latency
Chatter of applications on a LAN seldom matters due to the low
amount of latency. Normally, a typical LAN will not create more
than 1MS (milliseconds) of latency on your roundtrips. This means
that as a host sends a packet to another host on the LAN, it should
not take more than 1MS to get a response.
Latency comes from various sources, the first one being the very
hardware (switches, routers) that you are dealing with as well as
the media that you are using. Latency is increased by the distance
that a packet needs to travel from point A to point B, and it is
limited by the speed of light. Here is a summary of the different
sources of latency on a network:
- Propagation: This is simply the time it takes
for a packet to travel between one place and another at the speed
of light.
- Transmission: The medium itself (whether
optical fiber, wireless, or some other) introduces some delay. The
size of the packet introduces delay in a round trip since a larger
packet will take longer to receive and return than a short
one.
- Routers and other processing: Each gateway
node takes time to examine and possibly change the header in a
packet (for example, changing the hop count in the time to live
field).
- Other computer and storage delays: Within
networks at each end of the journey, a packet may be subject to
storage and hard disk access delays at intermediate devices such as
switches and bridges. (In backbone statistics, however, this kind
of latency is probably not considered.)
Look at a packet movement between hosts A and B on the same
segment of a LAN.
This being very fast, a normal user would never notice the
impact of TCP sessions and all the background messaging taking
place. For example, a Wireshark packet trace revealed that when a
user opens a simple Word document residing on a Common Internet
File Sharing protocol (CIFS) share on the network, over 1000
messages are generated between the two hosts. This is due to the
semantics of the application (Word) on top of the CIFS. Once again,
it's not an issue over the LAN, but try this over a WAN connection,
and you may have some problems.
For a corporate link between sites, latency is a reality and is
unavoidable. Regardless of the service level agreement that you may
have with the Telco (even if you are the Telco), the laws of
physics still apply.
Latency can be roughly calculated by the distance covered,
multiplied by the speed of light. This will give you the absolute
lowest number for the latency of your connection. You then have to
factor in all the devices introducing latency on the path such as
routers, switches, modems, etc.
Here Is an Example
ACME Company has an office in New York City and another in San
Francisco. Both sites may have some servers and ACME is leasing a
T-3 (44.736 Mbits/s). This is roughly half the speed of a 100MB
connection that a desktop user has in the office. However, users
are reporting that opening files that reside in New York from San
Francisco is really slow. Why? Latency.
