The Internet of 2025: its Future and Emerging Technologies

1
The Internet of 2025 its Future and Emerging
Technologies

• Scott Kirkpatrick
• Hebrew University
• Engineering and Computer Science
• Jerusalem, Israel

2
What is our vision for 20 yrs from now?

• Start with a vision and a problem to be solved
• Example Vannevar Bushs memex
• Problem dealing with the accelerating
  expansion of scientific knowledge
• Vision technology to capture data,
  new structures of linkage to navigate
• Things missed
• networks
• consequences of digital nature
  of all accessible information
• But this was in 1945!
• Today
• Same problem, not just scientific knowledge,
  but all consumer activities, all scholarly
  activities, perhaps the nature of government

3
To deliver results for 20 yrs from now?

• Key tasks for a distributed, ubiquitous global
  computing environment
• Fine-grain distributed search and retrieval
  of information
• (ultimate Google)
• Dynamic, adaptive, fine-grained provisioning
  of data
• and services
• (ultimate Akamai)
• All management tasks automated, all systems
  self-healing
• The technology vision global access of high
  bandwidth
• Embedding of high computing power in everyday
  things
• The Moores Law exponents double through
  expansion
• at both ends of the scale

4
Technology Starting Points

• Overlayer networks, with network links over
  todays backbone are already the basis for
  Internet growth
• All control algorithms are
  message-passing, but decisions need to be much
  more highly distributed for effective scaling.
• Exploit todays leading edge distributed
  technology prototypes as appropriate building
  blocks
• GRID protocols for service access on demand
  products now avail.
• Ocean Store for distributed retrieval
• Peer-to-peer is still mostly a concept, but
  essential
• hashing is a promising technology basis
  for distributed addressing

5
Where do we go next?

• Where do the three different types of networks
  occur?
• Grids dense computing, storage, and
  communications fabrics
• Scale-free the Internet and things in it
• Random graphs problems derived from other
  networks
• How big is the Internet, its information space,
  its underpinnings?
• At least 20 TB, but gt 100 TB of non
  robot-accessible content
• No one search engine covers all the accessible
  material
• How fast is it growing?
• Still doubling every year, changing at least
  every two months, but not observed on any shorter
  timescales.

6
Where do we go next?

• Many algorithms are developing in a way that
  makes them distributable
• E.g. survey propagation, belief propagation,
  turbo decoding
• What are the most important issues in managing
  it, or better, in managing organisms that live in
  it and grow with it?
• Things fail in place leaving a family of
  percolation problems
• Recovery speed more important than mean time to
  fail
• You cant optimize a constantly evolving
  organism, but you can regulate its growth
• Secret weapons are there effects of RSB in the
  Web?

7
Technologies from complex systems

• Open systems have emergent properties, seen in
  simulation
• Growth dynamics
• Phase transitions
• Good measurement of real network structure and
  dynamics needed
• Message passing is a way of life in distributed
  systems
• Market mechanisms can now be characterized
• Advances in optimization must translate into
  successful control
• Constraint satisfaction
• Game theoretical payoff replaces conservative
  interactions
• Flow control needs better physical understanding

8
Technologies from complex systems

• Structure emerges among peers
• Biological metaphors
• Sociological metaphors
• Economic/political metaphors
• Each must decrease overall cost, increase
  survivability
• Network introspection
• Imbed cognitive capabilities in the network
  elements
• Locally perform experiments to measure the
  environment
• Integrate the local view by message-passing

9
The challenges in getting ready for 2025

• Advances in algorithms, and in their application
• Measurement is vital
• Our models are proliferating, actual tests very
  rare
• Prototyping must address extremes of scale,
  realities of networked interation
• Necessary for adequacy of approaches
• Necessary to ever transfer ideas to real future
  implementations.
• Necessary to navigate the standards pathways that
  shape the future of data networks.