Transcript Group A - Princeton EDGE Lab

NITRD
Complex Engineered Networks
Breakout Session on Internet
Robert Doverspike (AT&T Labs – Research)
Keith Ross (Polytechnic Institute of NYU)
Sept 20-21, 2012
at&t
Internet Topic Areas for Questions and
Challenges
I.
Definitions
II. Security
III. Economics
IV. Privacy
V. Methodology
VI. Policy
VII.Reliability and performance, prediction
2
at&t
Definitions
Internet: Though different things to different people, we define 3
broad categories
• The public Internet. The original instantiation of the “Internet”
in the USA has evolved into a collection of independent,
commercial carriers who interconnect via pair-wise peering
agreements and protocols (such as BGP). This collection of
packet networks allows open IP addresses and reaches all
consumer and some business end users (eyeballs).
• Virtual Private Networks (VPNs). This category supports
independent business applications and enables private IP
addresses and various forms of privacy and security beyond the
public Internet. However, at some point packets of the public
internet and VPNs mix on the same links of the carriers.
• Private packet networks. These networks are heavily firewalled
(both physical and virtually) from the above networks and
usually built over a combination of separate routers/switches
and lower-layer infrastructures (such high-rate private lines,
multiplexing and cross-connect equipment, leased dark fiber, or
privately-owed fiber). Many mission-critical networks are of this
categories.
3
at&t
Security
• Henning! This needs to be rewritten to better focus the issues
and not make look like old laundry list Malware distribution is
perhaps the biggest challenge facing the Internet today (Bob
Kahn’s talk). What methodologies can be used to create a more
secure and private Internet?
• How dependent are various mission critical systems (such as
power grids, transportation system, network control planes, and
financial systems) on the various manifestations of the Internet
(Public Internet, VPN and private packet networks). What
should be done to better to secure the connections to these
systems, particularly in the context of cyber warfare. For
example, many of these forms of Internet have not proved to
be able to properly deliver large-scale synchronized data. Is
there a solution for this and what are the impacts to cost and
performance.
4
at&t
Economics
5
•
While critical questions and challenges (such as security, functionality,
characterization of the Internet as a system, QoS, measurement of its
health and performance) need to be explored, because of the
independent, commercial architecture into which the Internet has
evolved, these questions cannot be analyzed disjointly from economic
models, such as cost constraints and impacts and financial health of the
major ISPs who constitute the Internet.
•
We need to better understand the economic incentives that spur
innovation, such opportunities for computer networking innovation, visa-vis and the assessment of impact of disruptive architectures, such as
pathways to implementation and complexity of management,
operational procedures. Also, how do we gracefully evolve to new
architectures, given embedded infrastructures and reliability objectives
•
The cost and revenue of the Internet is dominated by the last mile or
“access network” to get to the end users ("eyeballs”). The edge is
where the “action” is; applied all the way to the edge and coordinated
in a heterogeneous (end-to-end) way, plus support emerging end-user
needs in context-computing, and content based management.
•
What should be the role of emerging CoS (and associated QoS) as a
deployed technology in the commodity Internet vs. cost containment
objectives, pricing models (such as flat-rate pricing plans, as well as the
price structure for access that cuts through the tariffs constraints and
at&t
issues), and security issues.
Privacy
• What should individuals, online services, regulators, and
policymakers do to ensure our privacy is protected in this age
of Internet data? As more user information is tracked (often for
purposes of marketing and advertising), how do we prevent or
control data mining by third parties to profile individuals in
detail?
6
at&t
Methodology
• What are the similarities or differences between the types of
problems and economic constraints of power grids and those of
the Internet? Are there methodologies that apply similarities
and synergies for the research & engineering community to
exploit?
• How can the Internet carriers provide usage data, internal cost
models, traffic considerations, architectures, and
implementation constructions to the academic and govt
research community so that they can better participate in its
evolution and optimization?
• To make networks more cost effective, lower power
consumption and more reliable, future networks will likely be
more agile in adapting to exogenous traffic and dynamic
changes in network states. This will put tremendous stress on
the control plane. Not all network states can be sensed and
used for control in the complex network of the future. The
challenge is to make Network Management and Control (NM&C)
a science so the architecture can be optimized and prevent
NM&C from dominating the cost or limiting network
performance. An example is: what is the role of SDN and how
does this tradeoff with security?
7
at&t
Policy
• Reword A potential role for government and research
community is assess how the current decoupled structure of the
Internet (where each carrier optimizes its own subnetwork,
CoS/QoS, and cost structure, perhaps at the expense of other
carriers or their users) compares to a structure where the
Internet is optimized as a whole, i.e., from the overall societal
advantage. What is that role? How does this relate to
objectives/notions such as net neutrality and cyber security?
What should be the role of the government and Internet (in the
context of independent, multi-carrier, commercial architecture
into which it has evolved) with regard to social issues, such as
Adopt (broadband) America Fund for rural access. How we can
build-out and operate high capacity reliable networks at much
lower cost (esp. to rural) areas?
8
at&t
Reliability and performance, prediction
• Reword What objectives and approaches are used to provide
network reliability and performance for example: Lack of
exposure and control of independent peering/transit (including
pricing) agreements can lead to instability. How do we analyze
and try to solve this? (mitigating the impact of component
outages and network maintenance and upgrading) vs.
survivability (outages due to natural disasters, acts of war,
outside traffic impacts, such as peering changes leading to
surges, and deliberate attacks). E.g., We need to better
understand the propensity for cascading failures and now to
mitigate them.
9
at&t
Backup
10
•
Does the Internet need to be scrapped and re-invented? If so, what
are the realistic pathways to make that happen? Remove
•
One way to address the complexity of Internet it to better clarify its
policy objectives. Remove
•
How is the Internet equipped to handle new application paradigms,
such as cloud computing (large data centers), branch-cache servers,
p2p. Are there new drawbacks/threats? Remove
•
What are the benefits of machine-to-machine (device-to-device ?)
networking (at edge) and what are the risks? Remove
•
This needs more definition How to combine heterogeneous data
sources to make inferences about higher level phenomena Remove
•
While much of the discussion involves the large ISPs that constitute
the “Internet”, large content providers and private networks, such as
Google, Microsoft, Amazon, DISA, etc., have built their own
backbones over leased private lines or dark fiber. Although the lastmile networks remain a critical component of the Internet, will tier-1
networks become irrelevant or at least unimportant.
© 2012 AT&T Intellectual Property. All rights reserved.
Questions/Challenges
1. Economics While critical questions and challenges (such as security,
functionality, characterization of the Internet as a system, QoS,
measurement of its health and performance) need to be explored,
because of the independent, commercial architecture into which the
Internet has evolved, these questions cannot be analyzed disjointly
from economic models, such as cost constraints and impacts and
financial health of the major ISPs who constitute the Internet.
2. Economics We need to better understand the economic incentives that
spur innovation, such opportunities for computer networking
innovation, vis-a-vis and the assessment of impact of disruptive
architectures, such as pathways to implementation and complexity of
management, operational procedures. Also, how do we gracefully
evolve to new architectures, given embedded infrastructures and
reliability objectives
3. Privacy What should individuals, online services, regulators, and
policymakers do to ensure our privacy is protected in this age of
Internet data? As more user information is tracked (often for purposes
of marketing and advertising), how do we prevent or control data
mining by third parties to profile individuals in detail?
11
at&t
Questions/Challenges
5. Security Focus this question on private packet networks. What should
be done to better to secure the Internet’s connections to power grids,
transportation systems, and financial systems? Particularly in the
context of cyber warfare. The Internet has not proved to be able to
properly deliver large-scale synchronized data. Is there a solution for
this and what are the impacts to cost and performance
6. Economics The cost and revenue of the Internet is dominated by the
last mile or “access network” to get to the end users ("eyeballs”). The
edge is where the “action” is; applied all the way to the edge and
coordinated in a heterogeneous (end-to-end) way, plus support
emerging end-user needs in context-computing,] and content based
management.
12
at&t
Questions/Challenges
10.Economics What should be the role of emerging CoS (and associated
QoS) as a deployed technology in the commodity Internet vs. cost
containment objectives, pricing models (such as flat-rate pricing plans,
as well as the price structure for access that cuts through the tariffs
constraints and issues), and security issues.
11.Methodology What are the similarities or differences between the types
of problems and economic constraints of power grids and those of the
Internet? Are there methodologies that apply similarities and synergies
for the research & engineering community to exploit?
12.Policy Reword A potential role for government and research community
is assess how the current decoupled structure of the Internet (where
each carrier optimizes its own subnetwork, CoS/QoS, and cost
structure, perhaps at the expense of other carriers or their users)
compares to a structure where the Internet is optimized as a whole,
i.e., from the overall societal advantage. What is that role? How does
this relate to objectives/notions such as net neutrality and cyber
security? What should be the role of the government and Internet (in
the context of independent, multi-carrier, commercial architecture into
which it has evolved) with regard to social issues, such as Adopt
(broadband) America Fund for rural access. How we can build-out and
operate high capacity reliable networks at much lower cost (esp. to
rural) areas?
13
at&t
Questions/Challenges
13.Reliability and performance, prediction Reword What objectives and
approaches are used to provide network reliability and performance for
example: Lack of exposure and control of independent peering/transit
(including pricing) agreements can lead to instability. How do we
analyze and try to solve this? (mitigating the impact of component
outages and network maintenance and upgrading) vs. survivability
(outages due to natural disasters, acts of war, outside traffic impacts,
such as peering changes leading to surges, and deliberate attacks).
E.g., We need to better understand the propensity for cascading
failures and now to mitigate them.
14
at&t
Questions/Challenges
18.Methodology How can the Internet carriers provide usage data, internal
cost models, traffic considerations, architectures, and implementation
constructions to the academic and govt research community so that
they can better participate in its evolution and optimization?
19.Methodology Vincent and Muriel rewrite What is the role of SDN leading
to more complex/powerful users with more complex (contextdependent) demand. How does this trade-off with security mandate.
20.To make networks more cost effective, lower power consumption and
more reliable, futhre networks will likely be more agile in adaptiung to
exogenious traffic and dynamic changes in network states. This will put
tremendous stress on the control plane. Not all network states can be
sensed and used for control in theh complx network of the future. The
challenge is to make Network Mangement and Control (NM&C) a
science so the architecture can be optimized and not make NM&C
eoither comainted the cost or limit network performance.
15
at&t
16
© 2012 AT&T Intellectual Property. All rights reserved.