A Swarming Architecture is Good for Internet Data Transfer

1
A Swarming Architecture is Good for Internet Data
Transfer ?
Offensed by Jiazhen Chen Alexander Kiaie
2
Question1 Why we need this?

• The Goal of your proposal is
• Use Swarm for Universal Data Transfer
• It is Better Because?
• Robust
• Multipoint-to-point transport layer
• Can solve the P2P problem
• A Headache for Network Administrator
• However, what about the other aspects?
• Potential Traffic Congestion
• Cost of numerous connections among peers
• Therefore, your swarm does not look like a
  proper pill for curing the headache.

3
Lets continue

• Despite your ambitious goals
• What is your motivation for this novel design?
• In other words
• Why we need lots of swarm to replace the current
  architecture?
• Possible Key Benefits
• Post-Popularity
• What about using Google Search SourceForge (OR)
• SkyDrive MSN Search
• If its legal, there is no barrier for storage
  and sharing online.
• Block availability
• Why, Why, Why we need to exchange blocks that
  might never trigger our interest ever???

4
Here we go

• The Comparison over Uswarm and isolated swarms
• You claimed Uswarm has potential benefit over
  the other
• But you admit that here not account for the
  selfish behaviour.
• Maybe they both perform equally bad )
• You also admit Its over-estimate the
  improvement in reliability since we did not
  precisely model the tracking process.
• So.. The tracking process is arbitrarily defined
  and simulated?
• The impact of DHT, Replicated Trackers on tracker
  availability
• It might be true - experiment
• BUT, in this proposal,
• No explanation on how you measure the result
• Thus, the conclusion is doubtable to us.

5
Its not over yet

• The incentive Strategies
• You said the cost is proportional to the number
  of open connections
• What if when two peers are willing to upload
  blocks but do not have blocks of interest to the
  other?
• The TCP inefficiencies, traffic shaping might be
  the cause of the low utilization of bandwidth in
  practice, not mainly caused by your focus
  problem.
• Without a long-term wide-covered test, it is hard
  to estimate whether your proposal can really work
  better than the current architecture.

6
Deployment

• Massive Replication
• Knowledge of routing topology
• Difficulty
• Peer knowledge
• Limit free-riding?

7
Other Work

• DHT
• Beehive, CoDoNS
• They work and have great properties

8
Plan of Work

• 1 Year for analyzing single-swarm behavior?
• Security and Privacy testing