As
part of the IT Architecture Initiative, the Office of Information Technologies
(OIT) is producing a series of papers outlining directions in information
technology architecture.
In the
spirit of RFCs, the papers are intended to facilitate understanding
of and open dialogue about information technology trends at Cornell,
with the ultimate goal of improving the utilization and interoperability
of information technology services throughout Cornell.
pdf version
Cornell
Data Networking: Wired vs. Wireless
Prepared by R.
David Vernon
SYNOPSIS
This document
outlines OIT recommended directions for "wireless" data networking at
Cornell. It includes:
•
An overview
of wireless networking.
•
Reviews
of relevant wireless technologies.
•
A description
of wireless implementation issues.
•
A brief
review of wireless security issues.
•
Recommendations
for wireless deployment at Cornell.
•
Practical
implications.
What
is Wireless Networking?
Generally
wireless data networks use the electromagnetic
spectrum (radio, micro & light waves) to carry data. Wireless networks
carry data from transmitters and receivers attached to computers to
fixed transmitters and receivers connected to the campus network infrastructure
via devices known as wireless access points. Access points are placed
at locations dictated by coverage needs and the nature of the signal
requirements of a given wireless technology. While some wireless applications
are focused point-to-point connections, others provide a general area
of coverage.
The advantage
of wireless data networking is readily apparent. Wireless data network
users are not confined to the locations of "wired" data jacks, and so
enjoy connectivity that is less restrictive and therefore well suited
to meet the needs of today's mobile faculty, staff and students.
An important
form of wireless data networking to be outlined in this paper is known
as cellular. Cellular service consists of many radio cells coordinated
with each other. In the ideal case, users of a cellular wireless data
network service can move from cell to cell throughout the coverage area
without degradation of performance or loss of service. The advantage
of a cellular approach is the ability to cost effectively tune radio
coverage over time to meet changing demands. This "cellular" approach
to service provision has been made familiar through mobile "cell phone"
services leveraged throughout the world.
Wireless
Network Deployment Technologies
Wireless
network deployment issues can be best understood if placed into the
following three service classifications:
- Wireless
Personal Area Networking (WPAN)
- Wireless
Local or Campus Area Networking (WLAN)
- Wireless
Wide Area Networking (WWAN)
Today,
the core technology behind the wireless service in each of these service
classifications is unique and more importantly, not an inherently integrated
seamless networking strategy. For example a user of a PDA, such as a
PALM (XXX) connecting to the Internet via a "Wide Area Network" service
provider will not be able to directly connect to Local or Campus area
wireless service currently being implemented at Cornell. Simply stated,
they are different services, with different hardware requirements, and
have fundamentally different service limitations.
WPAN Technology
WPAN systems
have evolved from "cord" replacement technologies. Some examples are:
- Cordless
communication between your keyboard and computer.
- Cordless
communication between your Personal Productivity Device (PDA) and
your computer.
- Cordless
communication within your home between your cell phone and your home
phone.
Because
of their initial function focus, WPAN wireless implementations to date
have been low-powered and offer limited coverage range.
The most
hyped of all WPAN wireless technologies today is called "Bluetooth."
Bluetooth is a product of the telecommunications and computer industry
"Bluetooth Special Interest Group"
and is rapidly gaining wide acceptance throughout the industry.
Bluetooth
is a low power wireless system initially designed as a supplement to
infrared-based cable replacement communications systems that have been
used for years. The capabilities are:
- Data
transmission speeds of 732 Kbps or less.
- Range
limited to ~ 10 meters.
- No
support for application roaming. (The ability to maintain connection
to the network when walking from one Bluetooth cell into another.)
Though
not initially a standard as defined by the IEEE, the IEEE now has a
formalized standards development in process for Bluetooth known as:
802.15.1. In turn, the IEEE is exploring the enhancement of 802.15.1
with a high data rate Bluetooth standard: 802.15.3.
A high power / high-speed version of Bluetooth may someday be seen as
a viable WLAN service.
WLAN Technology
WLAN systems
are designed to supplement and in some cases replace traditional wired-based
Local Area Networks. The predominant standards-based WLAN technology
being deployed in the United States and at Cornell today is based on
the IEEE 802.11b standard. (802.11b is the "upgrade" to the prior 2
Mb IEEE 802.11 standard.) The capabilities are:
- Transmission
speeds up to 11 Mbs.
- Target
range is ~ 30 meters.
- Limited
support for roaming.
Supported
implementation topologies include overlapping cells and integration
of remote cells via an existing network infrastructure.
The next
generation 802.11 wireless standards under development are:
- 802.11g:
Near term 20 Mb enhancement to 802.11b. Like 802.11b and Bluetooth,
802.11g broadcasts on the 2.4 GHz radio frequency.
- 802.11a:
54 Mb. Based on 5 GHz frequency. Not downwardly compatible with 802.11b/g.
Likely available within 2 years, but high initial component cost is
expected and deployments may require denser access point placement
for equivalent coverage of an 802.11b/g systems due to the nature
of 5 GHz vs. 2.4 GHz radio signal propagation.
One disadvantage
of 802.11 systems is their relatively high power consumption need. Given
this power requirement, many believe relative battery size/life requirements
of 802.11 services make this an impractical technology to support small
PDA class devices.
Competing
WLAN/WPAN technologies of note: HiperLan2
and HomeRF.
802.11x
and Bluetooth are not the only widely deployed and/or planned wireless
systems. Hiperlan2 is a 5 Ghz high-speed wireless standard under consideration
that has many compelling features. Included among these are:
- Automatic
AP configuration.
- Differentiated
service levels to support limited QoS.
- A
strategy for integration into WPAN / WWAN environments.
HomeRF,
a 2.4 Mhz system, also has a strong following. It is a WPAN "priced"
service with many features needed in small WLAN environments.
All that
can be said at this time about these systems is that the "popular wisdom"
today suggests these strategies will lose out to retooled Bluetooth
(802.15.x) and 802.11x systems over time. Regardless, it is worth maintaining
a watchful eye on these systems in case of a change in industry and
patron convictions.
WWAN Technology
(Cellular)
Wireless
phone corporations are using a broad range of technologies to transmit
voice and data in the Untied States and even more technologies are in
use across the globe. "Second generation" (2G) wireless Cellular phone
systems deployed in the United States today include those based on:
"GSM," "CDMA," and "TDMA" technologies. All these systems are currently
limited to datarates of ~ 9.6 Kbps. Within the next few years industry
plans are to supplement and/or replace these with second generation
"upgrades" to support data rates between ~ 170 Kbps to ~ 300 Kbps. Over
the next 5 years third generation (3G) standards-based solutions supporting
data rates in excess of 2 Mbs are expected (Gartner Group Research Brief:
Personal to Global: Wireless Technologies, 2005-2010.).
While
the pervasive nature of WWAN systems are enticing relative to the needs
of small mobile devices such as PDA's, the broader utility of the systems
remains limited. Clearly the available data rate of the provisioned
cellular infrastructure today is not sufficient to provide network services
to traditional laptop and desktop computers. In fact it is so limited
that cellular phone companies have developed esoteric protocols and
related gateway services, to compress traditional web information so
it can better be delivered over their cell networks.
The Wireless
Access Protocol (WAP) Forum is the major industry consortium. WAP
is a compiled and compressed protocol distilled from source Wireless
Markup Language (WML) content. (WML is a standards spin-off of Extended
Markup Language (XML).) Simply put, WML is similar to HTML but enables
the design of Web pages that can be displayed on the small view screens
of phones and PDA's. In turn, WMA is compiled by WAP "gateways" where
it is greatly compressed into the WAP protocol and then transmitted
across the wireless network to WAP enabled clients. All this processing
is done simply to allow the transmission of information over a very
low bandwidth wireless pipe.
There
is a huge trend in the wireless industry to support WAP. Understanding
the ramifications of this direction is important for Cornell departments
interested in extending information across the cellular phone network
to connected devices. The prime issue deserving thought is the design
of web content conforming to WML structure in parallel with standard
HTML, SHTML, and XML pages. Then, web
portal technology must be deployed to automatically detect the nature
of the requesting client and direct the appropriately formatted page
in response.
Satellite
Based Wireless Systems
Although
they are wireless, satellite based wireless systems are outside the
practical intent of this document to some degree. Nevertheless, satellite
based systems can be viewed as an augmentation to WWAN services. However
the cost value of service is often limited to specific applications,
most notably those applications that benefit from one-direction broadcasting
of content or communication to locations on the globe not serviced by
other means. Bi-direction communication using satellites to support
highspeed and/or pervasive connectivity on the Cornell campus simply
is not a practical approach today. But the use of fixed location satellite
receiving stations on the Cornell campus to receive broadcast information
from content providers is an important resource to leverage.
Wireless
Implementation Issues
There
are several key points to keep in mind when outlining the deployment
of wireless services.
- Because
of the "churn" in the wireless industry, any system deployed
today has a limited "state of the art" life span.
- Each
of the different types of wireless technology is targeted at a slightly
different user base and/or coverage targets.
- Inherent
interoperability of WPAN, WLAN, and WWAN systems is non-existent today.
- Each
type of system has limits.
In short,
the vision of a ubiquitous and integrated wireless service for Cornell
comprised of a selected suite of solutions is not likely in the near
term.
Wireless
also has limitations when compared to the wired network infrastructure.
There tends to be a fair amount of discussion of the potential for wireless
to replace all needs for wired a network at Cornell. This is not a viable
argument today. Wireless is simply no match for the high data speeds
(in excess of 1 Ggbs) and service quality levels deliverable on modern
wired networks.
The wireless
systems under consideration for Cornell currently enable shared 11 Mb
resource. (The 11 Mb rate is a function of the proximity of the receiving
client: the weaker the signal, the slower the connection.) Even with
expected enhancements to wireless data rates, wireless systems will
not have the same quality of service, or network application potential
of wired systems. To place this into practical context, wireless data
rate services are, at best, the equivalent of 10-year old twisted pair
wired networking technologies.
Other
wireless limitations include:
- No
current provisions for quality of service, thus eliminating support
for voice and video services(IEEE 802.11e work group is defining 802.11
QoS standards.).
- Interference
prone nature of the radio spectrum makes guaranteed service problematic.
- Short
"state of the art" technology life spans.
Regardless,
users of 11Mb wireless services world wide find excellent utility
within current data rate limitations - and wireless allows a wonderful
form of connectivity that wired systems can not afford – mobility.
Today, wireless networking has become a desired service class enhancement,
to the broader wired network at Cornell.
If one
of the great values of wireless is client mobility, deploying many uncoordinated
"one-off" implementations may be shortsighted. Enabling a
larger campus-wide service, however, is far more challenging. Challenges
include:
- IP
number allocation to mobile devices.
- Coordination
of wireless network deployment.
IP number
allocation is complicated by the mobile nature of wireless users. IP
numbers at Cornell are mostly bound to a physical location. (An IP number
is the unique Internet Identity of your host, not unlike the phone number
of your phone.)
In most
standard routed network configurations, you would not be able to unplug
your computer, walk to another building on campus, plug it in, and expect
the IP number to work. You would be on a different part of the network
and each part only supports a certain specific list of IP numbers (a
subnet).
Dynamic
Host Configuration Protocol (DHCP), can solve the problem of having
a static IP number assigned to your computer, but DHCP servers still
need to allocate a number that will work within the subnet of the computerís
given connection. To work around this issue for "mobile" computers requesting
IP numbers from DHCP servers, organizations have two options. One is
to leverage the installed campus routersí ability to create a campus
wide "virtual" network (VLAN) and forge a single dedicated network and
broadcast domain for projected wireless usage, thus allowing IP numbers
to work regardless of the router hardware they ultimately pass through.
Another approach is to consider the implementation of Mobile
IP.
Interference
caused by other devices using the same part of the radio spectrum is
another issue of concern for wireless network systems. 802.11b/g services
use the same spectrum allocation as Bluetooth, many cordless phones,
and other wireless devices. Users of wireless service should keep this
in mind when setting their expectations for service quality. Additionally,
wireless networks based on 802.11x must be coordinated. Two or more
access points need to work in concert with each other. While 802.11b/g
has 11 channels, cell to cell coordination requires at least 3 and popular
deployment wisdom recommends an unused channel between each. In short,
all 11 channels can be consumed by one integrated wireless system. In
an attempt to assure reasonable wireless service quality many peer universities
have strong central control over the use of the electromagnetic spectrum
and related wireless deployments. Many universities have policies banning
independent deployments and specific types of "offending" technologies.
Given Cornellís independent spirit, the notion of central control of
wireless air space may be distasteful, however lack of coordination
would be far worse. It will remain an important IT architecture goal
to figure out the best means to coordinate and resolve these conflicts.
The simple desire to work together may foster the greatest returns.
Wireless
and Security
Unlike
data on wires, which can remain inaccessible in walls, barred underground,
and locked behind doors; data carried on general coverage wireless networks
is there for all to receive. Open access to wireless data means that
if a transmission is of a sensitive nature, precautions must be taken
to protect the information. Most wireless technologies offer data encryption
to protect data from being read by an unintended recipient. However,
users of wireless networks should not assume all configurations have
enabled or effective security services.
Currently
interoperable "standards based" 802.11 systems, as defined by the Wireless
Electronic Computability Alliance (W-Fi) requires support for Wired
Equivalent Privacy (WEP) encryption services. However WEP is of little
to no value in large open deployments as is envisioned for Cornell.
WEP uses static "keys" that any user of the wireless system must know.
Once a master key code is disseminated to a large campus, it is silly
to assume any value in encrypting streams that anyone with the key can
simply decrypt. In turn, many pockets of small wireless nets, each with
their own "keys," would not allow roaming. This WEP limitation and other
sever security weakness of the 802.11 security services are well understood
by the 802.11 development community and hardware manufactures. The IEEE
802.11e (task group E) is in the progress of defining enhanced 802.11
security tools. However it is not clear when these will be available
or declared a universal requirement of inter operable 802.11 wireless
deployments as certified by Wi-Fi. In the interim individual vendors
are beginning to market vendor specific enhancements to their 802.11
offerings. The disadvantage of this approach is the requirement of a
sole manufacture source infrastructure to take advantage of the products
enhanced security features.
Wireless
and Authentication
Eavesdropping
is possible on a wireless network which means precautions must be taken
to prohibit unauthorized use of the service. In order to restrict illicit
consumption, additional authentication procedures are required. Most
are based on procedures requiring the pre-registration of legitimate
802.11 clients. A more sophisticated approach could be via an authentication
server that challenged user access before data packets could be transmitted
to the general campus resource. Not only is authentication an issue
relative to limiting access to the net, it is also important for users
of a wireless network to be assured the access point to which they are
connecting is really a legitimate agent of the expected service
provider. It is conceivable that anyone with a current 802.11 access
point and relevant hardware could easily "spoof" the legitimate campus
network and those devices within! As with the limitations with basic
WEP, the need for broader 802.11 security resources is being addressed
through the 802.11e process and one off vendor solutions.
Deployment
Directions at Cornell
Keeping
in mind the technologies and limitations outlined above, Cornell must
make a balanced investment in wireless services. Because wireless is
a new network utility, it also represents a new expense. Wireless deployments
will not lower the support costs of the wired system, nor will wireless
systems be able to deliver all the network services available on wired
networks. Given this, initial wireless deployments at Cornell should
be targeted where wireless utility is seen to have the greatest value,
i.e., at those locations where patrons will utilize en masse insecure
networking to support mobile computing devices. These locations
are likely to be locations such as University Libraries, dinning halls,
and other finite locations conducive to wirelessís value.
Ongoing
and recommended installations of wireless services at Cornell are based
on the WLAN technology 802.11b/g. After much practical review of WPAN/WLAN
development in the United States, 802.11x is the mainstream "commodity"
product of choice with a growing base of industry providers. But again,
by implementing one form of wireless you do not enable all classes of
wireless services. Users of an 802.11b enabled laptop computer will
not be able to use that same wireless system as their cell phone. Or,
users of a Bluetooth or WAP enabled palm will not be able to directly
connect with the 802.11b network.
Until
there is a convergence in wireless strategies, or effective gatewaying
tools, a unified vision of wireless is not a reality. This is not to
say the Cornell should ignore the expected increase in Bluetooth, WAP
enabled devices, and future high-speed 3G cellular services or the infrastructure
required to support them. But the value of the service must be balanced
against the cost. Cornell should keep abreast of ongoing development
to assure due diligence investment in new wireless systems at optimal
times.
Regarding
IP numbers, the current plan at CIT is to implement a DHCP service to
allocate IP numbers to registered hardware. This is the simplest and
most straightforward means to resolve IP allocation today. But it will
not resolve the long term desire to permit Cornell IP number allocation
to wireless devices outside of Cornell's wireless system, and it may
be problematic to provide service to tens of thousands concurrent users.
Practical
Implications
Departments
interested in 802.11b wireless services should be aware of the limitations
and unique benefits when considering deployment. In review these are:
Current Limitations and Deployment notes:
- 802.11
is not a Bluetooth or a cellular phone class service.
- 802.11
will not provide the same data rates and quality of service as a wired
network.
- Current
"public" 802.11x transmissions at Cornell are not secure!
- High
power consumption wireless means larger battery needs for many mobile
devices(There is conflicting opinion on the practicality of 802.11
for PDA class devices.).
- Wireless
systems are prone to interference. Be prepared for frustrated users.
- Wireless
systems must be coordinated if patrons expect campus wide mobility.
- Wireless
systems require a network configuration to support a roaming IP number.
- Future
wireless security systems being outlined by the IEEE 802.11e work
group will require campus wide coordination to be effective.
- Churn
in the wireless industry means demands for better/new wireless services
in the near term.
Current
Benefits:
- Mobility.
- Extending
network to areas that are cost prohibitive to run wire.
- Current
802.11b deployments will have valued utility for an extended period
despite new "state of the art" wireless developments.
Perceptions
of cost savings that may be afforded by wireless systems to extend network
connectivity to locations that are cost prohibitive for traditional
wired services must be realistically tempered. First it is important
to note that the users of WLAN's will still use the campus backbone
and Internet resources just as much as those connected via traditional
wired network. In turn, wireless hubs do not remove the obligation of
users to pay their fair share for the cost associated with the larger
campus and Internet services leveraged.
On the
surface, using wireless hubs to deliver network services to currently
wired stationary hosts may seem to be a way to reduce the total cost
of networking. In reality, replacing existing wired repeaters with wireless
only lowers the network quality to your patrons and reduces the networkís
ability to support enhanced services such as video on demand or video
conferencing.
Closing
Thoughts
Despite
limitations, the benefits of a coordinated, focused deployment of a
wireless network at Cornell is resolutely encouraged. The benefit of
allowing user mobility and extending the research and teaching value
of network accessible information to new locations is seen to have clear
value by the Office of Information Technology.
Ideally,
departments will see the inherent value of a coordinated implementation
ó an implementation that enhances mobility, the primary value of the
technology. In fact, there is a defensible argument to be made that
in order to take advantage of future enhancements in wireless technology
that will afford a secure and mobile wireless resource, deployments
must be coordinated. Minimally this would require:
- Campus
wide VLAN and DHCP services for wireless.
- Campus
wide coordination of radio spectrum allocation.
- Common
security and authentication strategy.
In parallel
with this initial deployment, Cornell Information Technologies and interested
departments must keep a vigilant eye on industry trends and be prepared
to retrofit the wireless system when appropriate. The long-term goal
is to create an integrated, seamless wireless system capable of supporting
the full range of network devices in use or envisioned.
Return
to Papers Page
