|
From a consumer’s perspective, the entire world is at their fingertips and
available through their phones, PDAs or laptops. It’s as if any piece of
information or content can be created, sent or received instantaneously. Videos
are uploaded or downloaded with a single touch and images and files shared
seamlessly. Any call or text is effortlessly completed with a touch or two on
their keypad and their “connectedness” to family, friends or business
associates is assured.
But for the telecom service provider, delivering this high level of service
and quality of experience is anything but effortless. Today’s service providers
manage networks that require tremendous amounts of hands-on involvement by
human operators. While current OSS’s and BSS’s provide a measure of automation
for some portions of a service provider’s operations, today’s network
management remains mostly an ad hoc effort. All the elaborate processes and
efforts to plan, configure, operate, manage, maintain and tune network systems
– with human involvement – accounts for a significant fraction of an service
providers annual operating expenses.
Some of today’s networks aren’t fully optimized for capacity or performance
and with the ramping increase in broadband traffic, operating costs continue to
soar. The forecast is that the pace of automation is being outstripped by the
growth of the network scale, capacity and capability. To support continuous
broadband traffic growth and new broadband service launches, telecom service
providers need flexible, scalable and efficient solutions that make network
build-out, configuration and service innovation and delivery simpler, faster,
more automated and less costly. At the core of the network, these
solutions must cost effectively contend with more bandwidth–intensive
applications and meet end-user quality of service expectations.
To breakout of incremental improvements to network management, Bell Labs
research is looking for solutions that increase efficiencies and reduce
operating expenses by orders of magnitude. Researchers look across several
management layers - service, network, element and network element - to
determine where the greatest efficiencies can be achieved. Rather than enhance
or add new management policies and processes that would add complexity and
increase human oversight - and additional cost – to network management, it
seems that the opposite approach may provide a solution. That is, allow the
network to manage itself.
|
|
“The scale and complexity of modern communication systems and networks,
and the progressively decentralized nature of their ownership are strong
reasons to look at self-organization and self-optimization as possible models
for management and control of these highly complex systems. Complexity here is
not merely a computational and transport challenge that could be overcome
through acquisition of vast amount of computational, switching and transmission
resources, it is the diversity of applications, volume of connections,
geographic spread of users, localized ownership of the network and
‘connectivity, anytime, anywhere’ with ever increasing bandwidth that make the
underlying systems challenging to manage in traditional ways. Even a 10%
reduction in network OpEx, easily expected from a self-optimized or Self-X
approach to network operations, translates to $10Bs of saving annually for
operators across the globe.”
- Iraj Saniee, Head, Math of Networks
and Communications Department
|
One novel approach to machine to machine interaction in future networks is
to adapt and leverage the control principles found in nature. Spontaneous
magnetization, crystallization, lasers and superconductivity are examples of
self-organization in physics where cohesive behavior emerges from initial
disorder. In self-assembly and auto-catalytic networks in chemistry, molecules
organize themselves in well-ordered arrangements without external action and in
biology we observe highly complex coordinated action as in the folding of
proteins, homeostasis and flocking. An example from physiology that is
even closer to networks is the self-optimizing and self-management phenomenon
found in the human body - the autonomic nervous system (ANS). The
ANS is a regulatory branch of a person’s central nervous system that helps them
automatically adapt to changes in their environment. The ANS helps regulate the
heart, breathing, temperature, blood vessels' size and blood pressure,
bronchial diameter (and thus air flow) in the lungs, stomach, intestine and
salivary glands, etc. All of this is achieved automatically without direct
conscious control. Blood is pumped, breaths are taken and food is digested and
we never think twice about it.
Communications networks can benefit in a similar manner, if the piece-parts
of the network - routers, switches, base stations, and numerous other
less-known elements all perform basic network operations – resource management,
traffic management, fault management and recovery, service upgrades, security
and others without resorting to human operator. For example, instead of
pre-planned partitioning of the wireless spectrum as it was done for GSM (and
may be needed for LTE), could base-stations self-determine their portion of the
spectrum as needed based on actual loads and negotiate resources with their
neighbors dynamically? It turns out that there are formal ways to perform
these tasks with neither massive signaling for coordination nor reliance on
human operators, just as in the above examples from nature.
Iraj Saniee has been leading an effort within Bell Labs Research known as
Self-X to derive formal mechanisms to enable self-organizing networks.
Pulling in distinct efforts and appropriate expertise from across Bell Labs
(including Murray Hill, Villarceaux, Stuttgart and Dublin), this team has been
able to define a series of solutions that progressively improve
self-sufficiency of networks. Formal techniques, such as gradient
methods, stochastic approximation, annealing randomization and particle
systems, are leveraged and combined with engineering insight to devise novel
Self-X algorithms with little communication overhead.
In a cellular network for example, having fixed or static frequency
assignment works well when the load on the network is uniform. However,
such “averaged out” load hardly ever occurs in practice. There are load
fluctuations all the time, which is better suited to dynamic spectrum
allocation. But even more so, when a portion of the network becomes
overloaded due to an unexpected event (e.g. a sporting event), or excessive
load due to a new technology (e.g. a cluster of wireless, high bandwidth
on-line gaming activity) there is no capability to automatically adjust to
conditions that are overwhelming the network. The result is degraded or
suspended service. To hedge against these, more spectrum and capacity needs to
be allocated upfront than needed, whereas a self-stabilizing dynamic allocation
would allow the cellular network to optimize spectrum by redistributing the
load and reconfiguring bandwidth and power allocation per cell.
Similar examples arise in parameter tuning in wireless networks, bandwidth
reservation in broadband access, and learning of dynamic configurations in
wireline networks. Decentralized control and a self-learning solutions
that operate throughout the network – with little intra-network signaling –
allows for higher level of flexibility and adaptability to meet to broadband
traffic growth and end-user QoS demands while keeping operator costs in
check.
|
“One can look at the handoff threshold in cellular networks as a function
that could be improved through an automated approach. Handoff threshholds are
currently fixed to predetermined operator-set levels but with a Self-X
solution, the cells in the wireless network would have added functionality to
learn and adapt to traffic and load to automatically and dynamically set the
threshold in order to minimize call drops. This then becomes an automated,
hands-off task handled within the operator’s network and improves the end-users
experience by delivering uninterrupted service.
- Iraj Saniee, Head, Math of Networks and Communications
Department
|
According to consumer trends, it is a natural development for end-users in
the future to have even higher expectations regarding seamless delivery and QoS
for their ever increasing high-bandwidth services. Through the examples
provided by nature, Bell Lab’s approach in developing this novel solution for a
self-optimizing network - Self-X – will help to not only meet their customer
expectations by automating and improving overall network operations, but will
significantly reduce their addressable OpEx as well.
Many of the potential advantages of Self-X are already anticipated by
industry standards. The 3GPP SON committee has been developing standards
for signaling and data between network elements in anticipation of
self-organizing in wireless LTE networks. However, challenges remain to
demonstrate to traditional network managers that a Self-X type of approach is
effective and practical. The expectation is that the LTE standard will
facilitate the introduction of Self-X in communication and telecommunications
in the next 2-3 years.
|
