Transcript Horizon2020 Consultation Meeting (Smart Networks & Architectures

Future Network Technologies Research and
Innovation in HORIZON2020
Consultation Workshop
Smart networks & architectures Panel Session
Brussels (29/06/2012)
[email protected]
http://www.ictconsulting.ch/presentations/horizon2020.ppt
29/6/2012
Olivier Martin
1
Outline

The “Invisible” or “infrastructural” Internet


The OTT, aka the “Services” Internet




Facebook, Google, LinkedIn, Netflix, Skype, Twitter, YouTube, etc.
IPV6 Deployment Status & Issues
The proliferation of standards-making organisations


The Tiered model
The fading role of the IETF
What can Horizon2020 do?
Tentative Conclusions
29/6/2012
Olivier Martin
2
State of the Internet (1)

What I used to write:


There are two Internets that have very little in common, namely:
 Academic & Research Internet (GEANT & National Research and
Education Networks (NRENs) in Europe, Internet2 & National Lambda
Rail (NLR) in the USA, etc.)
 Commercial, also dubbed “Commodity”, Internet, itself divided between
the wired, wireless and cellular Internets.
However, they clearly participate to the Global Internet despite major
differences. Indeed, the A&R Internet has the following distinguishing
characteristics:




29/6/2012
Bandwidth rich unlike the bandwidth scarce Commercial Internet
Special emphasis on interconnection between Universities and Research
centers as well as provision of dedicated infrastructure for “missionoriented” communities (e.g. CERN/LHC)
Provision of advanced core Internet services (e.g. IPv6, native Multicast)
(too) Strong emphasis on bandwidth on demand. Was there a solid
business case analysis made?
Olivier Martin
3
State of the Internet (2)

Another formulation, probably closer to reality, is that there are indeed two
Internets “piled” on top of each other, the “infrastructural” one that is largely
“invisible” and is having troubles and the “services” one that is flourishing,
more specifically:

The “Invisible” Internet (provided by Telco, specified by the IETF) plagued
by:





The lack of inter-domain services (QoS, Multicast)
The endless transition to IPv6
Near exponential growth, change of usage patterns (e.g. video), Falling
revenues, Lack of spectrum.
Hence the “legitimate” temptation to add more and more functionality (e.g.
compute & storage)
The Over The Top (OTT) Internet



29/6/2012
What uneducated users believe to be THE Internet
Unconstrained by standards
Driven by creativity and business opportunities
Olivier Martin
4
The Ubiquitous Internet
Old, politically incorrect, slide without Africa, MiddleEast, Russia and former USSR Republics and others
The «Tiered» Internet
29/6/2012
Olivier Martin
6
OTT Providers & Services

Successfully provided what the Internet was unable to deliver, namely:







Quality of Service by moving content closer to the user
Inter-provider services because of its “inherent” Overlay nature
Application-level/Emulated Multicast
Sophisticated Peer-to-Peer technology (e.g. BitTorrent)
Smart applications (e.g. Facebook, Google, LinkedIn, Netflix, Skype, Twitter,
YouTube, etc.)
Last but not least, Cloud services of various kinds (i.e. SaaS, PaaS, IaaS, etc.)
So, do we need really need to change anything and move from a “relatively
dumb” network to a “smarter” network, being well understood that end-user
devices as well as services will become “smarter and smarter”?

Given the outstanding success of OTT services the answer is a resounding NO, I
think!
29/6/2012
Olivier Martin
7
The “ill-cooked” and outdated IPv6
migration strategy (1)

Started in the early 1990s, in the days of class-based addresses, i.e.
before Classless Internet Domain Routing (CIDR) came about, in
parallel with the development of IP next generation (IPng) proposals
that finally led to the adoptionIPv6:




Exhaustion of IPv4 addresses was seen as imminent
The Internet was very small compared to today!
Dual-stack was thought to be straightforward, which it actually was
not as, around the same time, Digital Equipment Corporation (DEC)
dropped DECNET Phase V, that was based on the same migration
principles, on the grounds that it would never fly!
Lack of variable length IPv6 addresses, i.e. lack of backward
compatibility with IPv4, was probably the biggest mistake:

However, back in the early 1990s, this was deemed impossible by the
VLSI designers
29/6/2012
Olivier Martin
8
The “ill-cooked” and outdated IPv6
migration strategy (2)

Back in 1999 during the Internet Architecture Board (IAB) workshop in
Utrecht, the idea of taking advantage of the “dreaded” Network Address
Translators (NATs) to facilitate the transition to IPv6 was coined by Gabriel
Montenegro (SUN) who was also working on an alternative to IPv6 dubbed
”Realm Specific IPs” (RSIP) that was not pursued.


However, the IAB was violently against NATs because of the violation of the endto-end paradigm. In other words, NATs were then “evils”!
Indeed, the IETF only standardized NATs and associated Application Level
gateways (ALGs) long after they had been widely deployed by commercial ISPs.

So, nothing happened for 10 years, i.e. 2009 until it became obvious that, if
only because of the upcoming lack of IPv4 addresses, the IPv4 and IPv6
Internets would need to coexist for many years, if not decades!

Having acknowledged the new reality, the new challenge, therefore, was to ensure
“graceful” interconnection between these two Internets such that they would appear
as a single, “Global” Internet
29/6/2012
Olivier Martin
9
Various statements about NATs and IPv6-IPv4
coexistence

Russ Housley (IETF Chair): Are NATs for IPv6 a necessary evil?


“They are necessary for a smooth migration from IPv4 to IPv6 so that the
important properties of the Internet are preserved….We need to be pragmatic”
Xing Li (co-author of the IVI proposal (RFC6219) also known as
“stateless NAT64”)
 “The experience with IPv6 deployment in the past 10 years indicate that the ability
to communicate between IPv4 and IPv6 address families would be beneficial.”

Network World 20/3/09


“Business incentives are completely lacking today for upgrading to IPv6, the next generation
Internet protocol, according to a survey of network operators conducted by the Internet Society
(ISOC).”
Special Network World Issue 21/1/09 (sponsored by NTT)

IPv6: Not If, When?
29/6/2012
Olivier Martin
10
IPv6 launch events: An amazing but
genuine recurring series of “flops” (1)

Started with the Global IPv6 launch event in Brussels in January 2004




Most impressive event, all major actors present
Many misleading, actually false, statements about the new capabilities
inherent to IPv6 like built-in Quality of Service!
Fully Internet-enabled car demonstrated with only one major feature
missing (driverless)!
Continued by the IETF in March 2008: “The night the IETF turned off IPv4”

Discovered a few “residual” problems!
 “There's still a lot to do before the IPv4 addresses are used up within
4 years or so!”
 We are now exactly there, and we are definitely much more
advanced, although actual deployment of IPv6 is despairingly slow!
29/6/2012
Olivier Martin
11
IPv6 launch events: An amazing but real
recurring series of “flops” (2)

Pursued by ISOC in 2011 & 2012 “World IPv6 Day”


Lot of “hype” before, essentially no or very little feedback afterwards!
Conclusions and Observations





Relative fiasco in terms of wider and faster adoption of IPv6
Intoxication does not work, the Internet is driven by economic
considerations not by ideology and/or proselytism!
Genuine advances have been made with the provision of new IPv6/IPv4
transition tools
An IPv4 trading market is developing which is not a very good sign
However, IPv6 appears to be unavoidable
29/6/2012
Olivier Martin
12
Will IPv6 be deployed soon?

We are now mid-2012 and although IP6/IPv4 transition technology has
become available, e.g. Stateful and Stateless NAT64, it is not widely deployed
in operational networks yet:




Not clear how well NAT64 technology works and whether it will scale!
Fortunately, most popular servers are already dual-stack or will soon be, however,
many hyperlinks will probably appear as “broken” depending on the access
network.
Carrier-grade IPv4 NATs are being deployed which implies NATs over NATs (i.e.
NAT444) which may, in turn, increase the number of “broken” applications and
induce developers to layer new applications on top of the Web (i.e. HTTP)
In any case, it is very comforting to observe that the IETF clearly learned the
lesson and that there is now an intensive activity on NATs of all kinds!

NAT44, NAT64, NAT66, NAT444, … let us hope we will not see NAT4444 or even
worse “tricks`!
29/6/2012
Olivier Martin
13
IP4-IPv6 Internets coexistence



NAT64 coupled with DNS64 should allow IPv6 only hosts to access the
IPv4 Internet using “synthetized” IPv4 addresses and a stateful
gateway.
However, as pictured in the next slide, the reverse is NOT possible in
an automated manner, therefore manual configuration is needed in
order to connect IPv4 networks to the IPv6 Internet which is very
impractical on a large scale.
Whatever happens with IPv6, i.e. fast or slow deployment, it is too
late to have dual-stack hosts everywhere, hence there will be IPv6only islands that will not be able to “easily” communicate with IPv4only islands and vice-versa, hence:

Enabling “minimum” interoperability between these two worlds is a
MUST, even if the functionality is limited (e.g. Web, e-mail)
29/6/2012
Olivier Martin
14
NAT64 Technology: Connecting IPv6 and
IPv4 networks
(Cisco Whitepaper - April 2012)
29/6/2012
Olivier Martin
15
The ambiguities of the Standards making process




Excerpts from Larry Roberts’ 1995 article “The ARPANET & Computer Networks” :
“With five, independent, public packet networks under construction in the 1974-1975
period (USA, Canada , U.K. , France , Japan), there was strong incentive for the nations
to agree on a standard user interface to the networks so that host computers would not
have unique interfacing jobs in each country. Unlike most standards activities, where
there is almost no incentive to compromise and agree, carriers in separate countries can
only benefit from the adoption of a standard since it facilitates network interconnection
and permits easier user attachment”.
As pointed out by L. Roberts unless that there are compelling reasons to find a common
solution diverging commercial interests often lead The new rule seems to be slowdown and complicate the standardization process rather than to accelerating it.
Good Standards are the ones that are used and do not over or under specify.
Most standards have a very limited lifetime, however, the IPv4 protocol suite and
CCITT/ITU transmission standards are notable exceptions to this general rule.
29/6/2012
Olivier Martin
16
Internet Engineering Task Force (IETF)

IETF’s track record has been rather disappointing during the last 10
years or so, however, it remains the undisputed guardian of Internet
protocols:



Failed to adapt in a timely manner to the rejection of its dual-stack IPv6
migration strategy
Failed to keep its ability to swiftly define suitable standards
Defined MPLS, the last major architectural Internet change




However, the Internet community is divided about the wisdom of this very
significant, also very complex, change
Interactions with ITU are less than satisfactory (MPLS-TP, NGN)
Is the much heralded “rough consensus” model still working?
The ongoing development of IP4-IPv6 transition/coexistence tools is, in
my view, very encouraging

29/6/2012
However, Internet purists are probably in violent disagreement
Olivier Martin
17
Standards Developing Organisations (SDO)
A general observation is that there are far too many Internet related SDOs with
the result that the Internet entropy is increasing slowly but surely.

IETF is not the only culprit, the track record of the W3C (Web Consortium)
and the Open Grid Forum (OGF) is also disappointing in terms of practical
impact but not in terms of number of “standards” produced!

The Open Networking Foundation (ONF) that is dealing with Software
Defined Networking (SDN), aka OpenFlow, is anything but open, i.e. the entry
price is high.

The same can be said of MIT’s Communications Future Program (CFP)

ITU rates well with respect to transmission technology and Optical Transport
Network (OTN)





However, where is ITU’s NGN and MPLS-TP leading to?
Interactions with IETF leave much to be desired.
IEEE, being more “industry-driven” has, by far, the best track record with
Ethernet and WiMax
But, ETSI also fares very well with GSM and LTE standards.
3GPP is linked to ETSI and ITU with very focused goals
29/6/2012
Olivier Martin
18
What the Internet may look like in the future


A “Green”, i.e. energy aware, Internet will appear.
Broadband access (i.e. Mb/sGb/s) will become nearly ubiquitous



Wireless access will become prevalent (3G, 4G, LTE, WiMAX, 5G)
But, fixed access will not disappear (ADSL, FTTH, GPON, Cable TV, leased
lines, etc.)
Paradigm changes are unavoidable, e.g.:






Host based  Content based
Information Centric Networks (ICN)
Peer-2-Peer networks (P2P)
Content Distribution Networks (CDN)
Software Defined Networking (SDN, i.e. OpenFlow in practice)
LISP "Locator/ID Separation Protocol (LISP)" (RFC6115)
29/6/2012
Olivier Martin
19
Challenges/Observations (1)



Keep the “Global Internet” connected.
The existing “dumb” network model that allowed the emergence of
OTT services is probably better suited to innovation and fast creation
of new services than a functionally richer network encompassing
additional functionality such as compute & storage.
The “fading” role of the IETF is preoccupying:


the Internet belongs to everyone but who owns and controls it?
The “multiplication” of new standards making organisations/forums is
also worrying (e.g., W3C, OGF, ONF).



Remember that Web 2.0 did not originate from W3C
Likewise, the Cloud emerged because of the inability of the OGF to produce
adequate standards in a timely manner.
OpenFlow/SDN brings bright/promising ideas, however, it should not be used as a
pretext to delay the introduction of IPv6
29/6/2012
Olivier Martin
20
Challenges/Observations(2)


Some new proposals, e.g. “bandwidth on demand” or “all-optical” networks
look like “solutions looking for a problem” and lack a solid business case
analysis.
The Internet entropy is increasing, which body is really controlling its
development:




The past 10 years have been disappointing to say the least
The last major architectural change was MPLS, however, the Internet community
is still spit/divided about it
The early Internet was extremely primitive in terms of services (i.e. ftp,
telnet, email) before the advent of the Web (HTTP & HTML) that did not
really originate from the IETF.
The Internet can be visualized vertically as Telcos, ISPs, OTTs, Users on top of
each others:

Lack of concertation with users is flagrant
29/6/2012
Olivier Martin
21
Challenges/Observations(3)



The OTT allowed what the IETF completely failed to provide, namely:
Inter-provider/inter-domain services, QoS, Multicast
Although there is a clear trend towards multiple networks (e.g. IoT,
cars, smart electricity Grids) this should not dilute the effort to keep a
single Internet even though, it is already fragmented because of its
multilingual nature and the increasing number of paying services.
The top priority is to improve the connectivity between the IPv4-only
and IPv6-only worlds:


NAT64/DNS64 is a right step in that direction but additional “middleboxes”/network appliances are likely to be needed.
In principle, a promising market, given the need to ensure minimum
interoperability between these two worlds at the Web and email level, at the very
least!
29/6/2012
Olivier Martin
22
Additional comments (1)




The trend is likely to be multiple, loosely coupled, networks rather than one
single network (e.g. smartGrids, home networks, VANs,IoT, sensor nets), if
only for cyber-security reasons.
A well-organized transition to IPv6 should resemble more the ongoing
transition from analog to digital TV than the complete “mess” surrounding
the IPv4/IPv6 transition!
In-network storage looks, a priori, like a good idea in terms of optimization,
however, is it worth the additional complexity and related CAPEX/OPEX costs
(e.g. router memory is likely to be an order of magnitude more expensive
than regular RAMs).
Many other parallel Internets will appear, e.g. the Internet of Things (IoT),
Smart Grids, etc., that, if only for cyber-security reasons, will only be loosely
connected to the “Global” Internet through highly-secured gateways.

Will they be based on IPv4 or IPv6 or some other loosely related technology?
29/6/2012
Olivier Martin
23
Additional comments (2)

Related to the likely proliferation of Internets that will be dealing with objects
of various kinds, one could say that there will be two categories of Internets:


The “human” Internet, also known as the “Global” Internet, which human beings will use to
connect smartphones, tablets, notebooks, ultrabooks, laptop PCs) in order to access and use
social networks, Web services, more generally OTT services, including “cloud” services.
The “non-human” Internet, e.g. the IoT, various sensor networks, that will be used to collect
and aggregate control information to be fed ultimately into the “Global” Internet for actions
by humans or into “specialized” Internets (i.e. control networks) for action by robots.
29/6/2012
Olivier Martin
24
What can/should Europe do in
Horizon2020 (1)

Despite my rather negative comments about the IETF, this is the ONLY forum
dealing with Internet architecture and lower layers protocols (i.e. up to layer
4 (transport), therefore Europe:





Must definitely strive to strengthen its presence and influence
Try to establish strategic trans-continental alliances (e.g. EU-US, EU-JP or EU-KR) at the
academic and research, project, and, most importantly, industrial levels.
Must forget about European-only solutions in a more and more globalized world
Must look at the wider worldwide market not only to the intra-European market which is huge
but too fragmented.
Standards making organizations must carefully avoid to over specify standards, in order to
allow innovation while still allowing, either interworking or achieveng particular goals (e.g.
CO2 emission levels, electricity consumption)
29/6/2012
Olivier Martin
25
What can/should Europe do in
Horizon2020 (2)

Opportunities for the European ICT industry are definitely huge:






irresistible trends towards a “Greener” ICT world
almost borderless room for innovation in the OTT service space, e.g. e-Health, Smart-*
Cloud computing
Commodity routers, Large scale NATs , innovative network appliances facilitating
interconnections with the emerging IPv6 world
mobile Internet that is still in its infancy with immature protocols and products,
innovation in efficient content distribution (Information-Centric Networking, improved peerto-peer schemes).
29/6/2012
Olivier Martin
26
Tentative Conclusions


The IPv4 Internet is growing fast but cannot continue “as is” beyond 2014 or
so!
IPv6 looks “almost” unavoidable but is by no means “guaranteed” to happen!




IPv6 by itself only solves ONE problem, i.e. the lack of addresses BUT nothing else
Last major, but still controversial, architecture change was the introduction of
MPLS
The “Global” Internet will encompass two Internets, IPv4 and IPv6 with some
“glue” between them in the form of “dual-stack” hosts (mostly servers).
Many other parallel Internets will appear, e.g. the Internet of Things (IoT),
Smart Grids, etc., that, if only for cyber-security reasons, will only be loosely
connected to the “Global” Internet through highly-secured gateways.

Will they be based on IPv4 or IPv6 or some other loosely related technology?
29/6/2012
Olivier Martin
27
Additional Slides
29/6/2012
Olivier Martin
28
The Internet and NGN
(Tomonori Aoyama - NICT)
29/6/2012
Olivier Martin
29
A New Generation Network
– Beyond NGN –
(Tomonori Aoyama - NICT)
29/6/2012
Olivier Martin
30