Postcards%20from%20the%20Edge%20

1
Postcards from the Edge Offense

• Presented by
• Sam McIngvale
• Rob Kotz

2
General Premise

• Why Cache-N-Forward?
• The authors argue that advances in wireless
  technology will render the current Internet
  architecture useless.
• How?
• What makes new mobile devices unstable?
• TCP has been effective thus farwhy change?

3
General Premise v2.0

• What are the benefits of Cache-N-Forward?
• How will changing the current architecture fix
  the unstability?
• Will there be an increase in speed?
• Will there be an increase in throughput?
• What will be different?
• Why should I want to change?

4
General Premise v3.0

• Paper argues the need to push complexity into
  the network and make the end-to-end transport
  protocol simpler.
• This conjecture contradicts the fundamental
  principle the original Internet was built-on.
• Keeping the network simple has gotten us this
  farwhy change?
• The authors offer no reasoning to support where
  this conjecture comes from.

5
General Premise v4.0

• Is Cache-N-Forward a new architecture?
• All that is happening is the file is getting
  moved somewhere closer to the requesting
  end-user.
• Then the end-user receives the file via a
  TCP-like protocol.
• Should we implement a new architecture that only
  improves large file-transfer?
• Real-time applications are increasing and
  user-experience (QoS) is a major concern.

6
Overhead

• How much overhead will Cache-n-Forwarding
  introduce into the Internet?
• We dont know because the authors dont tell us.
• Reading/Writing files to disk at each hop will
  dramatically slow-down the rate of transmission.
• Disk I/O is already the limiting factor for many
  operations.

7
Overhead v2.0

• For a single mobile-device (end-user) to request
  a file, the following steps must take place
• The receiver contacts a file name resolution
  service to resolve the location of the requested
  file.
• The receiver looks-up the host in DNS.
• The receiver sends a query to the sender.
• The sender contacts a name resolution service
  that resolves the name of the mobile host to a
  set of PO nodes.
• The sender forwards the file to one or more POs.
• POs hold the file until contacted by end-user to
  arrange delivery.
• Finally, there is delivery of the file via
  direct-transmission (TCP?).
• I would much rather just take my file directly.

8
Overhead v3.0

• This architecture is not appropriate for
  real-time applications.
• Definitely does not solve any issues with QoS a
  major concern for future architectures.
• All we get is a guarantee of delivery (already
  present with TCP) but we sacrifice extreme
  transfer times.

9
Current State of Proposal

• The authors recognize several key questions that
  still remain with their architecture.
• How should routing be integrated with caching?
• How do we implement congestion control?
• What are the storage needs of a CNF node?
• What are the necessary post office descriptors?
• How do we specify routing tables?
• How should cache-based multicast be implemented?
• Why I am I writing this paper?
• What is the Internet?

10
Current State of Proposal v2.0

• This is an Internet Architecture proposal.
• You CANNOT leave important issues such as routing
  tables and congestion control up in the air.
• We would like to argue specific (low-level)
  design and implementation issues, but the design
  is so high-level there is nothing to argue about.

11
Current State of Proposal v3.0

• The Wireless Revolution the authors are hoping to
  address is happening right now.
• By the time their architecture is anywhere near a
  state that can be deployed, this issue will have
  been solved or addressed (if it needs to be).

12
Evaluation?

• Instead of offering concrete results to support
  their conjectures we are presented with the
  difficulties of testing such an architecture
• There are no concrete results to support the
  validation of this new architecture
• Its ok thoughwe have an Impact Statement and
  Intellectual Merit

13
Deployment?

• The authors offer no plan for deployment.
• They are proposing a major () overhaul of the
  Internet for this to ever happen there must be
• 1) Concrete results to justify the change.
• 2) Some plan to completely change the current
  infrastructure.
• Neither of these issues are addresses.

14
Conclusion

• Does not feel like a science paper.
• The authors make wild claims with absolutely no
  backing evidence.
• There is no evaluation of their proposed
  architecture.
• Seems this paper is more of a preliminary outline
  of a project yet to be completed.