Road to wireless Net has bumps
By Ed Acosta, President and CEO, BroadCloud
Communications Inc., Austin, Texas
EE Times
February 26, 2001 (12:03 p.m. EST)
While fewer than two percent of Internet users worldwide
have wireless access, expect that to change dramatically
over the next three years.But mobility will not come
without a fight. Before widespread acceptance and
availability, the wireless Internet must attain credibility,
which means significant advancements in reliability, speed
and bandwidth must be made before it is adopted broadly.
Many fundamental obstacles that compromise the promise
of the wireless Internet are rooted in the traditional
Internet Protocol suite, which was not designed for
wireless networks. Although Transport Control
Protocol/Internet Protocol (TCP/IP) was envisioned for
transmission across any physical channel, the traditional
IP suite-specifically, TCP-is built upon a set of
fundamental assumptions and optimizations specific to the
performance characteristics of wireline physical channels
available nearly 30 years ago. TCP is designed to
guarantee reliable transfer of data across networks
exhibiting packet loss in the range of 1 to 5 percent,
which is typical of wireline networking technologies.
Unfortunately, the 15- to 30-percent packet loss regularly
exhibited by wireless networks breaks the machinery of
TCP, pushing it into a zone in which the reliability it was
designed to ensure fails. The fact that any data transfer
can occur is a tribute to the resilience of the protocol,
albeit at the expense of much greater transmission time.
There are obvious performance differences between
wireless and wireline physical channels that are obvious in
the transport layer of a network stack. And there are
other factors that are dramatically different in wireless
networks, mainly higher bit error rates (BER), longer
latencies and highly variable channel availability. These
differences make channel conditions highly dynamic in
comparison with the relatively static channel conditions of
wireline networks.
Underlying assumptions
The tremendous variance in channel characteristics and
conditions created by these differences breaks the
underlying assumptions and optimizations upon which TCP
is built, causing its machinery to fail. TCP is, in effect,
thrashing in mobile cellular networks, throwing away much
of what precious little bandwidth is available. Also,
wireless networks typically use many different RF channels
to facilitate continuous communications; wireless devices
are forced to carry out a number of RF channel
bookkeeping and administrative tasks that temporarily
interfere with data communications. The flow control
mechanisms of TCP were not designed to maximize data
throughput in the presence of this activity.
These are but a small sampling of the obstacles that
interfere with the ability of the traditional Internet
Protocol suite to deliver reliable and efficient data
communications in mobile cellular networks.
The disparity in reliability and speed between wireline and
wireless networks created by these obstacles must be
rectified before the Internet, or the Internet as we know
it, can be successfully utilized in a wireless environment.
To date and for the foreseeable future including second-
and third-generation networks, wireline networks will offer
at least an order of magnitude greater capacity for data
throughput than wireless networks. Because of the
differences described above, it is not economically feasible
to provide the same level of data throughput in mobile
cellular wireless networks as is possible in wireline
networks.
Three fundamental matters adversely affect the
performance of any data networking system in a mobile
environment:
Physical channel performance of mobile wireless IP
networks is dramatically different and less efficient
than the performance of wireline networks.
More noise and distinctly different patterns of noise
adversely affect data communications, leading to
high BER and packet data loss;.
Wireless networks must employ frequency-agile
systems that utilize many physical RF channels
dynamically allocated to provide a logical channel
structure, thereby giving the appearance of a
dedicated physical channel. This will ensure the
economic feasibility of wireless networks.
Each presents unique challenges to data communication
systems that conform to the Open Systems
Interconnection networking model, such as the Internet.
In particular, at the transport layer, TCP is ill-equipped to
deal with the network performance that results from these
fundamental attributes of wireless networking and
performs poorly as a result. Key challenges traced to the
TCP design include its unreliability in wireless environments
and inefficient use of bandwidth.
The challenges of the RF environment push the packet
loss ratios in wireless networks beyond 5 percent, the
upper limit of the zone where TCP can guarantee reliable
transmission. At a 15 percent packet loss, within
specifications for normal operation in the cellular digital
packet data (CDPD) network, the probability of successful
transmission is 90 percent. The protocol will attempt to
compensate through multiple retransmission efforts, but
incur an expense of increased time.
Speed and reliability are not unrelated or mutually
exclusive benefits in TCP. Not only is file transmission
unreliable, but is slowed whether or not the transmission is
successful. This is because TCP is a retransmission and
acknowledgment-based protocol that compensates for a
lack of reliability and high BER by taking more time to
complete the transmission. Thus, as RF channel conditions
worsen and/or the mobile unit is in a saturated cell site,
transmission time increases at a greater-than-linear rate.
As the file size grows, transfer time degrades at a velocity
greater than the linear rate. Rather than functioning
smoothly, the machinery of TCP is thrashing, leading to
inefficient channel utilization that results in data
throughput efficiency's reaching only 25 to 35 percent vs.
a typical 75 to 80 percent for wireline networks.
Consequently, the perception of wireless' being slow is
caused by the lack of reliability brought on by trying to
use TCP in an environment for which it was never
designed. Moreover, if channel conditions get truly tough,
TCP just stops working.
The normal function of the Internet Protocol suite breaks
down in the wireless environment, leading to greater
transmission time and lower data-throughput rates. The
transmission time required increases greater than a rate
linear with file size. For example, in the CDPD network
once the file size exceeds 10 kbytes, data throughput
drops off exponentially to approximately 4 kbytes/second.
The situation is exacerbated by larger files. The same
effects occur in other network types such as General
Packet Radio Services. Engineers from leading wireless
companies including Ericsson and BroadCloud have
researched and documented this phenomenon. Initially,
the prospect of a wireless Internet generated a great deal
of excitement in the wireless and financial communities
and even with consumers.
But current offerings have fallen short of expectations and
have led to skepticism among potential users and
analysts. The promise of the wireless Internet will not be
met until the challenges of reliability, speed and bandwidth
are met.
The future for the mobile Internet is anything but dismal
and represents an enormous market opportunity. The
complete Internet, wirelessly, will be available sooner than
we might expect.
Most innovations will come from new companies that will
play a major role in defining the wireless Internet. New
products and services are concurrently being developed
and generally fall into one of four categories:
next-generation wireless networks, wireless and wireline
network infrastructure integration, mobile-commerce
applications and data compression.
Traditional telecommunications technology vendors are
working on advanced RF technologies driving
next-generation wireless networks-third generation, or 3G,
and beyond-and will be able to provide greater raw
bandwidth. Several new companies are concentrating on
the marriage of wireless networking technology and the
existing wireline Internet infrastructure. Other companies
are developing and deploying exciting mobile-commerce
applications. Lastly, companies are producing streaming
compression solutions to improve the speed of existing
wireless networks.
Developing innovations
Innovations concurrently being developed by leading
companies are certain to yield futuristic wireless networks,
network infrastructure integration, mobile-commerce
applications and advanced data compression. However,
the arrival of new wireless technologies will be largely
disconcerted, just as the development of new innovations
have been independent thus far.
New products and services necessitate a reliable wireless
platform. Without one on which to operate, they will not
deliver added value. Quite simply, the benefits of
innovations cannot be employed until there is a foundation
for them.
The creation of a reliable platform that powers new
applications and a patented alternative to TCP will enable
industry leaders to create an Internet that is completely
wireless.
BroadCloud has responded to the reliability challenge by
creating a wireless platform with its Wireless Internet
Protocol, an alternative protocol to TCP. Through internal
tests, BroadCloud's engineers have demonstrated that
Wireless Internet Protocol makes wireless networks twice
as reliable, up to 10 times faster and uses up to 75
percent less bandwidth.
The protocol client runs on a user's mobile device-a
personal digital assistant, pocket computer or laptop, for
example-and works as a transparent software background
that resides between the application layer and the
network stack. The client intercepts data passing from
applications to the network for transmission over the
wireless Internet and redirects the transmission to use the
more efficient Wireless Internet Protocol.
Data is then transmitted by the client to a corresponding
Wireless Internet Protocol server on the network. The
server then forwards the data to the final destination
using the original protocol specified by the application,
essentially conducting the data transaction with the end
destination so that neither the application nor the end
destination realizes that any intermediary was involved.
Moreover, during the process the protocol improves the
overall data exchange by optimizing various higher-level
protocols, such as HTTP, SMTP and POP3, and the
content they deliver.
Emerging networks
Emerging 2G+ and 3G wireless networks promise higher
data rates than those of today's networks. They will
deliver greater bandwidth within three to five years, but
what will this mean to the end user? Realistically, very
little.
That's because of two factors. First, commonly available
wireline bandwidth will improve significantly during this
time, but the size and nature of Internet content will
adjust accordingly. Second, based on industry forecasts,
both the number of wireless Internet users and their
demands on bandwidth will increase exponentially. A
greater number of users will share higher bandwidth
networks and as a result the average user will see little if
any improvement in performance.
Moreover, an enormous bandwidth gap between wireline
and wireless performance exists and will continue for the
foreseeable future. Users, influenced by their wireline
Internet experiences and expectations, will initially be
largely disappointed. |
