Skip to content

Instantly share code, notes, and snippets.

@Haseeb-Qureshi

Haseeb-Qureshi/network_and_consensus.md

Created October 10, 2018 22:19
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save Haseeb-Qureshi/abc212af5b420220450c462f7efee14e to your computer and use it in GitHub Desktop.
Save Haseeb-Qureshi/abc212af5b420220450c462f7efee14e to your computer and use it in GitHub Desktop.
Download ZIP
Scaling Blockchains at the Network and Consensus Layers at CESC
Raw
network_and_consensus.md

Scaling Blockchains at the Network and Consensus Layers

Emin Gun Sirer

The Promise

  • Blockchain is becoming a big deal.
  • Every finance company is developing a "blockchain strategy."
  • $5B invested in startups, $10B invested in ICOs.
  • Big gap between what's been sold to users, and what current technology is capable of.
  • Doesn't scale to being a payment system.
  • What about a store of value?
    • Consumes more than two nuclear power plant's worth of electricity
    • Alternatively, costs as much energy as the country of Denmark
    • $300M annually in value leaking out of Bitcoin
  • A decentralized system?
    • ~19 mining pools in BTC, ~11 in ETH, 21 BPs in EOS

Core problem

  • The systems we have cannot sale, cannot implement things like micropayments.
  • BTC is around 3 TPS, Ethereum about 10 TPS, credit cards around 5K TPS, Alibaba is around 325K TPS.
  • Scalability only gets worse as the system gets more popular.
    • Many bullshit techniques for artificially increasing throughput on a testnet
      • Bundling transactions, such as into 1TB blocks and making decisions on those
      • Centralizing through a single actor/machine/committee

Scaling at the Network Layer — The Forgotten Layer 0

  • All blockchains fundamentally depend on quick dissemination of data
  • The techniques they use for this are still rudimentary
    • Store and forward networks
    • Similar to primitive computer networking in the 1980s
  • BloXroute addresses scaling at the network layer
    • By disseminating information quickly from sources to sinks
    • Like Akamai, but for small, timely, high-value data with many sources and sinks
  • BloXroute democratizes the playing field for miners, stakers, and validators
    • Is a block distribution network, regardless of the underlying blockchain

How BloXroute works

  • Miners send and receive blocks to the BloXroute network through open source gateways
  • Uses 100x less data using internal caching
  • Propagates this data 10-100x faster using cut-through routing
  • Maintains neutrality via end-to-end encryption
  • As tokens use BloXroute to route blocks, BloXroute takes a small fee, and BloXroute's token entitles all token holders to a portion of those accrued fees

Scaling at the Consensus Layer

  • On May 19, 2018, a group calling itself Team Rocket dropped a PDF describing a new consensus protocol family
    • Called Avalanche
    • Achieves an entirely new type of consensus
  • Traditionally, two categories of consensus protocols:
    • Classical (Lamport-Liskov)
      • Quick finality, but does not scale
      • Quadratic message complexity
      • Precise membership, permissioned
    • Nakamoto Consensus
      • Robust, no need for precise membership
      • High latency
      • Low throughput
      • Does not scale
      • Wastes energy, not green, not sustainable

Avalanche family

  • Goals: best of all worlds
    • Quick finality, low latency (~2 seconds)
    • High throughput (1k - 10k TPS)
    • Scales (10K to 10M nodes)
    • Robust, no need for precise membership
    • No energy waste
  • Inspired by epidemic protocols and gossip networks
    • Security guarantees controlled by the system designer
    • Easy to understand the basis for security
  • New idea: deliberately metastable
    • Randomly sample a small number of peers, ask them their current state
    • If you observe a majority winner, change your state to that
    • Then re-sample again
    • Should take a constant number of re-sampling until the whole system converges
  • Properties:
    • Efficient and scalable
    • Can tolerate a large fraction of Byzantine participants, even with imprecise membership
    • Egalitarian ecosystem, no mining
    • Can participate via cell phone
    • Maximal concurrency—constructs a DAG in parallel

Avalanche (Ava) Platform

  • Multiple virtual machines
    • Ava, the native token
    • Additional tokens can be created
    • Ava-BTC, Ava-ETH, Ava-Zcash, etc
  • Sybil deterrence via stake
    • Staking nodes receive interest via minting
    • No slashing, no fund loss
  • Economic governance via voting
    • Can modify system parameters, within bounds
    • Crowd oracle for economic decisions, to sidestep central planning

Ava, native token

  • Serves as payment rail
  • People talk about "consensus" process in Bitcoin as if it were a way to sample opinions of participants
    • In Ava, it actually is
  • Transaction fees paid in Ava
  • Governed through repeated sampled voting
    • Ava interest rate for staking between 0-5%
  • Adding smart contracts layer, supports EVM-style smart contracts, 1000X better improvement over Ethereum

Performance

  • 7 TPS for Bitcoin, 15 TPS for Ethereum, 364 for Algorand, 1300 for Ava.

Summary

  • BloXroute optimizes and tokenizes data transmission on blockchains
  • Avalanche is a new platform that uses a novel consensus scheme to produce a more efficient, scalable, decentralized smart contract platform.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment