logos-storage-research/presentations/dagger-presentation.md
2021-05-13 15:39:42 +02:00

2.9 KiB

title theme revealOptions
Dagger solarized
controls progress transition
false false fade

Dagger

Decentralized Storage

Note: Briefly describe what Dagger is.


Why Dagger?

IPFS

Note: Decentralized storage is a fundamentally unsolved problem.
P2P storage hasn't taken off, why?
Without reliable storage, dapps won't take off.
Needs same guarantees as cloud storage, and then some.


Reliability

predictable

accessible

Note: Reliability is the property that is most lacking in p2p storage networks.
Predictable storage duration.
Storage is always accessible to you.


Proofs

data possession

retrievability

Note: How do we achieve reliability?
Proof of data possession over time
Proof of retrievability (interactive)


proof of data possession

Proof of storage 1

Note: Proof of storage.
File is divided into chunks.
Over time, proofs of different randomly selected chunks are requested


proof of data possession

Proof of storage 2

Note: We improve on this scheme by making sampling events dependent on each other.
This increases the odds of detecting missing chunks over time.


proof of retrievability

Proof of retrievability

Note: Prevents data withholding (ransomware)
The prover does not know when it's being verified


Incentives

storage

bandwidth

Costs of resources

Note: Incentives compensate for resources
Incentives provide security; it makes economic sense to stick to the rules
Storage incentives provide storage reliability
Bandwidth incentives provide retrievability (dynamic CDN)
Incentives encourage participation in the network


Money / Investments

Note: how much of the market can we cover
how much would it cost to make


Privacy & Anonymity

Privacy Stack

Note: Dagger layer: identities used for incentives and payments are pseudonymous.
Libp2p layer: peer identity is pseudonymous.
IP layer: depending on the need, there are several solutions to choose from


Network Overview

Network


Software Stack

Software Stack


Research agenda

  • proofs for large data sets
  • formal/emperical evidence
  • consensus engine

Note: Small proofs, constant size, aggregatable, computationally light
Zero knowledge proofs
Formal proofs/emperical evidence that our algorithms have the desired properties
Model the network to observe emergent properties
How does the storage network interact with a consensus engine