Joseph Baylock

1
From Silicon to AirwavesThe Long View of
Systems and Networking

• Joseph Baylock
• Martin Reynolds

These materials can be reproduced only with
Gartners official approval. Such approvals may
be requested via e-mail -- quote.requests_at_gartner.
com.
2
Our Predictions Through 2012

• The real-time enterprise is career-critical
• Moores law lives on!
• Quantum oddities become important
• Ultra-Wide Band (UWB) means wireless everywhere
• Lambdas reach to the server

3
Key Issues

• Will Moore's Law remain valid during the next ten
  years?
• How close to "free" will the various types of
  metro, long-haul and wireless bandwidth become?
• Where will the bottlenecks to a fully
  interconnected world appear?
• Will regulatory change (or the lack of it) stymie
  industry growth?

4
Real Time

• Computing power wherever, whenever
• Data access wherever, whenever
• Central surveillance of operations
• Real time makes cash
• IT challenge
• Harness technology to business goals
• A driver for new implementation

5
2000 Industry Size US Billions
E-Business
Telecom Services
Telecom Hardware
Services
Software
Hardware
0
200
400
600
800
1,000
6
2005 Industry Size US Billions
E-Business
Telecom Services
Telecom Hardware
Services
Software
Hardware
0
2,000
4,000
6,000
8,000
E-business will fuel the technology industry
7
The Separated Computing Model
Local GB
8
When Bandwidth Becomes Cheaper Than Computing ..
Death of Distance The speed of light
(propagation delay) becomes a concern Extremes
of distributed and centralized computing are the
norm, for example, networked and network-based
server mega-farms (grid computing) Move the data
to wherever there is (excess) computing power
Client/server, thin client, P2P computing are
the vogue Function-specific computing
Accelerators for XML, SSL, databases, etc.
9
Grid Computing

• Easy in theory, hard in commercial practice
• Security, discovery, authentication, chargeback
• Web services and grid standards are converging
• Latency
• Niche solution to current business problems
• Data mining a good candidate now
• New approaches to architecture needed
• ASP peak-load grid computing model could work

10
When Will Moores Law Expire?

• Helping
• EUV lithography to 2010
• SOI isolation
• Stretched silicon
• Isotopically pure silicon
• High-K dielectric
• Low-K dielectric
• Copper
• Stacking
• Moores law is safe from physics until at least
  27 July 2017
• But what about economics?

• Hindering
• Quantum leakage
• Power consumption
• Economics

11
Laws Were Made to Be Modified
Biological
Applications integrated
Photonic
Mechanical (MEMS)
Electrical Circuits
Years Away From Common Use
Moores Law maintains its currency and vibrancy
through integration of mechanical, optical and
biological functionality to control mirrors,
generate light and sense biological conditions,
etc.
12
Overall Wafer Fab Utilization
Ratio of Silicon Shipment to Fab Capacity
100
90
80
70
60
50
4Q97
2Q98
4Q98
2Q99
4Q99
2Q00
4Q00
2Q01
4Q01
2Q02
4Q02
2Q03
4Q03
9-02
13
Hardware Outpaces Software
MHz
Outlook and IE
Opportunity Gap
Windows
Windows gets efficient!
Windows 95
286 to 386
14
Application Possibilities

• So What?
• Videoconferencing
• Constant computing
• Still Doesnt Work
• Speech recognition
• Mobile sync.
• Single sign-on

• Usable and useful
• Ego-casting
• Video editing
• New ways to program
• Self-healing Web services
• Catalog and search
• New GUI
• Cameras everywhere

15
Quantum Effects

• Quantum mechanics is at the fringe of reality
• Strange behavior of subatomic particles
• Will result in many new technologies by 2012
• Quantum computers can factor large primes
• Breaks PKI technology
• But not in the foreseeable future
• Quantum cryptography emerging
• Could replace PKI
• Closer than quantum computers
• But how do you know it is working?

16
Bits and Qubits
17
A Qubit Register
18
The Networking Hype Cycle
Visibility
Ethernet MANs
Wi-Fi Ho Spots
Now - two years Two - five years More than five
years
l Services
Location-Aware Services
Softswitches
GMPLS
Enterprise CDN
DoS protection
Wireless Web
In the Cloud Security Services
NG Satellite
WiFi/802.11b
Optical Switching
DWDM
ATM
NAS
Video Streaming
IPv6
VoIP
Bluetooth
Managed Services
MPLS
ADSL
802.11a/g
ASPs
FTTH
GPRS
Intrusion Prevention
WLL
FC SANs
IPSEC
3G Network Capability
IP Telephony
UWB
iSCSI SANs
All Optical Network
Peak of Inflated Expectations
As of July 2002
Technology Trigger
Slope of Enlightenment
Plateau of Productivity
Trough of Disillusionment
Maturity
19
Network Security in 2012
2012
Today

• Security Platforms
• Incorporates firewalls, intrusion protection,
  IPSec VPNs, vulnerability assessment and gateway
  antiviral
• Speeds up to OC48 (at least)
• Firewall protection up to Layer 7
• Access
• Smartcard with biometrics
• Security Responsibility
• SP-built and operated
• Enterprise-managed

• Firewalls
• Protection up to Layer 4
• IDS
• Intrusion detection
• Access
• Password
• Security Responsibility
• Enterprise-built, operated and managed

20
Palladium

• A truly unique approach to platform security
• Can open up new business models
• Think about trust-based models now
• Expected in 2005, broad by 2008
• Will it work?
• Will it survive Hollywood?
• Will it be open?

21
Palladium System Architecture
Applications
Applications
Applications
Operating System
Operating System
Operating System
Secure API
Nexus
Nub
Nub
(Security Kernel)
(Security Kernel)
(Security Kernel)
Nub
Nub
Environment
Environment
Security
Agents
Security
Security
Agents
Agents
Secured
Secured
Secured
Unsecured Hardware
Unsecured Hardware
Unsecured Hardware
Hardware
Hardware
Hardware
Nexus
Environment
22
Ultra-Wide Band Wireless Everywhere

• Heavy on digital, light on analog
• Antenna spreads the spectrum
• 100 Mbps and 30 feet at milliwatt power
• Gigahertz of bandwidth
• Think wireless USB or Bluetooth 3
• Essentially free wireless for every chip
• Alternative Just wideband
• Requires scalable analog

Capacity and Range vs. Bandwidth
Air-Carrying Capacity vs. Bandwidth
23
Capacity and Range vs. Bandwidth
24
Air-Carrying Capacity vs. Bandwidth
3000
2500
2000
1500
1000
500
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Relative Bandwidth
25
Mesh Wireless Networks

• Broadcast Point-to-Point Cellular
  Microcellular Mesh
• Dense, high-bandwidth, low-power mobile nodes
• Battlefields and cell phones
• Broadband and thermostats

26
The Silicon Chronometer

• Nanosecond accuracy per week in isolation
• Translates to a foot a week
• New classes of cryptography
• Enables dynamic mesh networks
• MEMS then quantum
• True synchronous networks
• On your wrist in 2020

27
The Silicon Chronometer
28
Wireless Networking in 2012

• WLAN is nearly ubiquitous
• 3G actually arrives 4G is on the horizon
• Bluetooth painfully arrives
• Wireless is in (almost) every electronic device
  (up to three forms of it)
• Location-based services are valuable and usable
• The always-connected supranet arrives
• The detrimental social aspects of the always
  connected society are ignored blurring the line
  between work and personal time

29
Open Spectrum A New Paradigm for Radio Access
Paradigm Stop assigning and policing every slice
of radio spectrum and focus on intelligent
transmitters that can police themselves for
interference (the Ethernet model). Payment is
according to use. Pros Frees upfront private
capital to invest in wireless infrastructure Lower
s barriers to entry for smaller players spurring
innovation Current license holders could resell
excess capacity on the spot market In uncertain
times, auctions and licenses keep scarce spectrum
unused Cons Changing international spectrum
allocation is hard big changes are
harder Technical parameters (power, noise) can be
managed, but the billing algorithm can get
complicated Some users must be protected
30
What Is a Lambda?

• Italian car? Small sheep? Latino dance?
• No, a single wavelength of light
• 250 lambdas per fiber
• 10 Gbps per lambda
• Optical channels within a fiber
• But optically switched to make a network
• The next unit of purchase for bandwidth

31
Hybrid (OEO) vs. All Optical (OOO)

• Long-term economics of O/E conversion (when will
  it be pennies?)
• Migration of legacy metropolitan-area networks
• Do we route in the electrical or optical domain?
• Pace of progress on optical silicon (electrical
  and thermal-coupled switching)

32
Electrons or Photons?
33
The Optical Race Betting on Physics for Fun and
Profit

• Key Optical Switch Features
• Convert/modify lambdas
• Amplify/attenuate optical power
• Protection switching
• Multicast and add/drop (2-, 3- and 4-way
  junctions)
• Signalling protocol participation
• OSMINE compliance

34
Toward the All-Optical Network
Accelerators vs. Retardants
Improvements in semiconductor packaging and
cost Operational cost reductions - Carriers
believe now Cable industry moving toward telecom
(data) services - Municipal approval High
bandwidth services (SANs, lambdas-to-the-server
and inside-servers)
Routing algorithms must evolve (Cisco and
Juniper) Telecom bubble must wash out -
Chapter 11 is their friend Competition for RD
funds at major vendors Regulatory concerns,
special interests, consumer, government
35
Regulatory Impediments

• Hollywood
• Government surveillance taxes
• Corporate privacy
• RTE makes it worse
• Carnivore
• RBOCs and LD
• Bankruptcies

36
Who Pays for This?

Source 10K Filings
37
From Silicon to Airwaves The Long View of
Systems and Networking

• Joseph Baylock
• Martin Reynolds

These materials can be reproduced only with
Gartners official approval. Such approvals may
be requested via e-mail -- quote.requests_at_gartner.
com.