Over the next ten-to-fifteen years, it is anticipated that significant
qualitative changes to the Internet will be driven by the rapid proliferation
of mobile and wireless computing devices.Wireless devices on the Internet will
include laptop computers, personal digital assistants, cell phones, portable
media players, and so on, along with embedded sensors used to sense and control
real-world objects and events. As mobile computing devices and wireless sensors
are deployed in large numbers, the Internet will increasingly serve as the
interface between people moving around and the physical world that surrounds
them. Emerging capabilities for opportunistic collaboration with other people
nearby or for interacting with physical-world objects and machines via the
Internet will result in new applications that will influence the way people
live and work.
The potential impact of the future wireless Internet is very
significant because the network combines the power of cloud computation, search
engines, and databases in the background with the immediacy of information from
mobile users and sensors in the foreground. The data flows and interactions
between mobile users, sensors, and their computing support infrastructure are
clearly very different from that of today’s popular applications such as email,
instant messaging, or the World Wide Web.
Considering the wide range of future wireless network usage scenarios (4G
cellular/mobile, WLAN, mesh, P2P, DTN, sensor networks, vehicular networks,
sensor/pervasive systems), it is important to extract a set of common
requirements general enough to meet these needs, as well as those of future
applications that cannot easily be predicted today.
It is important to note that these requirements should apply to future
access networks and the Internet protocol stack as a whole in view of the
increasingly predominant role of wireless end-user devices. The current
approach of designing specialized networking solutions for cellular systems, ad
hoc nets, sensor applications, and so on leads to undesirable fragmentation
(and hence poor scalability, lack of interoperability, inefficiencies in
application development, etc.) among different parts of the network, and needs
to be replaced by a unified end-to-end protocol architecture that supports
emerging requirements of both wired and wireless networks.
To elaborate further, basic transport services of future Internet protocols
should reflect intrinsic radio properties such as spectrum use, mobility,
varying link quality, heterogeneous PHY, diversity/MIMO, multihop, multicast,
and so on, and the capabilities of emerging radio technologies such as LTE,
nextgeneration WLAN, Bluetooth, Zigbee, vehicular standards such as 802.11p,
and of course, cognitive software-defined radio (SDR). In addition, Internet
protocol service capabilities should be designed to serve emerging uses of
wireless technology, not only for conventional mobile communications, but also
for content delivery, cloud computing, sensing, M2M control, and various other
pervasive system applications. Below are some key requirements for future
networks designed to support the range of wireless usage scenarios;
1. Dynamic spectrum coordination capability: Historically, network
protocols have been designed to support resource management in terms of wired
network concepts such as link bandwidth and buffer storage. As radios become an
increasingly important part of the network, it will be useful to be able to
specify and control radio resources within the networking protocol itself. For
example, control protocols should be able to support dynamic assignment of
spectrum to avoid conflicts between multiple radio devices within the network.
Just as current IP networks incorporate protocols such as dynamic host control
protocol (DHCP) for address assignment, future networks could incorporate a
distributed repository of spectrum usage information that could then be used to
assign nonconflicting spectrum to radio devices when they join the network.
2. Dynamic mobility for end-users and routers: As more and more
end-user devices become wireless, networks will need to be designed to support
mobility as a normal mode of operation rather than as a special case. This
means that end-user devices should be able to attach to any point in the
network (i.e., global roaming), with the network providing for fast
authentication and address assignment at a very large scale. Currently
available mechanisms such as DHCP and mobile IP represent a first in this
direction, but a more general solution could involve a clean separation of
naming and addressing where each device would have a unique name, but would
only be assigned a routable address local to the network with which it is
currently associated (and this routable address may be as general as a
geographic location, i.e., geo-address). The main challenge is to provide a
distributed global name resolution and address assignment service that scales
to the level of billions of mobile devices. Because wireless devices may also
serve as routers in some of the ad hoc environments discussed earlier, the
network should be able to support dynamic migration of subnetworks. In
addition, dynamic handoff of traffic from one point of attachment to another
may be required for certain connection-oriented services.
3. Fast discovery and ad hoc routing: Because several wireless usage
scenarios involve ad hoc associations and continuously changing network
topology, it is important for the network to support fast discovery of
neighboring network elements. Discovery protocols for ad hoc networks should
support efficient topology formation in multihop wireless environments taking
into account both connectivity requirements and radio resources. Multihop
wireless scenarios further require efficient ad hoc routing between network
elements with dynamically changing topologies and radio link quality. The ad
hoc routing protocol used in wireless access networks should seamlessly
integrate with the global routing protocol used for end-to-end connectivity.
4. Cross-layer protocol stack for adaptive networks: Routing in
multihop wireless networks requires a greater awareness of radio link layer
parameters to achieve high network throughput and low delay. This means that
the network’s control plane should include information about radio link
parameters to be used for algorithms that support topology discovery and
routing. A key architectural issue is that of determining the appropriate
granularity and degree of aggregation with which this cross-layer information
is exchanged across different parts of the network (i.e., access, regional,
core, etc.).
5. Incentive mechanisms for cooperation: Ad hoc mobile networking
scenarios typically involve cooperation among independent wireless devices. It
will be important for future Internet protocols to include protocols that
enable such cooperation, first by advertising resources and capabilities to
neighboring radios and second by providing mechanisms for exchange of credits
or barter of resources in return for services such as relaying or multihop
forwarding.
6. Routing protocols for intermittent disconnection: Today’s
Internet routing and transport protocols are designed under the assumption of
continuous connectivity. However, this assumption is no longer valid for mobile
devices that frequently experience disconnection due to radio signal fading
and/or service unavailability. Future protocols should be designed for
robustness in presence of occasional disconnection. In order to achieve this,
the network generally needs to be able to store in-transit data during periods
of disruption, while forwarding messages opportunistically when a path becomes
available.
7. Transport protocols for time-varying link quality: Reliable
delivery of data on the Internet is currently accomplished using transport
control protocol (TCP) for end-to-end flow control and error control. TCP is
known to perform poorly in wireless access networks that are characterized by
higher packet error rates than wired links, along with time-varying bandwidth
caused by variations in radio channel quality and medium access control (MAC)
layer contention. Future transport layer protocols should be designed to work
efficiently in presence of packet errors and varying end-to-end bandwidth –
this will require the ability to distinguish between congestion in the network
and channel quality variations.
8. Efficient multicasting and multipath routing: The wireless
channel has inherent multicast capabilities, that is, a single packet sent by a
radio is simultaneously received by all receivers within the transmission
range. This property can be exploited to improve network performance in various
scenarios, but the routing and transport protocols have to be enhanced to
support multicast operation as a core capability. Radio multicast also opens up
the possibility of multipath routing in which multiple independent paths are
used for routing a single packet to improve end-to-end reliability and delay.
9. Location awareness and geographic routing: As discussed earlier,
emerging pervasive computing applications (i.e., vehicular, sensor, M2M) often
require the ability to delivery packets to an entire geographic region rather
than to a specific IP address. Also, for mobility services, knowledge of the
current geographic location is central to providing various new services such
as navigation and geographic search. This means that future networks should
provide location information as a basic control plane capability. In addition,
it would be desirable to optionally offer geographic multicast and routing
modes by which packets can be delivered to a specified geographic region.
10. Content- and context-awareness: A number of future network
service scenarios involve content addressability or content routing. For
example, an M2M application might involve a query for a particular
functionality (such as “printer”), and it would thus be useful if the network
protocol can resolve a content query to one or more specific network addresses.
Another network capability to be considered is that of content routing by which
network routers forward traffic based on content attributes of the data being
carried.
