Collection nodes (medium number, high-speed internet connection required)
Collection nodes are divided by the protocol into cooperating clusters and are assembling transactions into collections. They are verifying that transactions are valid to then forward commitments to collections to consensus nodes (stake in fixed increments). Each node in a cluster receives equal work and is compensated in proportion to their stake.
Consensus nodes (large number for decentralization, low hardware requirements)
Consensus nodes form and propose blocks, as well as ensure the security of the network. Consensus nodes can stake in any increment without seeing a disproportionate increase in work, which could otherwise make participation prohibitive (consensus is assigned on a per-vote basis). This is important to allow a large number of decentralized nodes in order to prevent corruption and increase resilience.
Execution nodes (small number of powerful machines/data-centers for scalability)
Execution nodes are executing transactions and maintaining the execution state. Execution nodes must stake in fixed increments with relatively high minimum stake, which should make it expensive for these nodes to fall offline. They receive ordered and finalized chunks of transactions from consensus nodes, and then request the underlying data from collection nodes (which get slashed if they cannot do so). Execution nodes ensure scalability of the network (and will further scale with Moore’s law).
Verification nodes (large number for decentralization, low hardware requirements)
Verification nodes are checking chunks, a fixed unit of work and must stake in fixed increments to ensure they are not assigned more than they can effectively handle (no maximum number of participants in the verification process: anyone can stake to join). Verification can be parallelized and thus can scale over a large number of weaker nodes compared to execution. The concept of specialized proofs of confidential knowledge are used to solve the verifiers dilemma (that nodes could actually not do the work of verification and still get the reward; see tech paper 3 for details). Observed faulty executions are reported to the consensus nodes (with a cryptographic proof), which adjudicate the received challenges and slash malicious actors. While the architecture can significantly increase throughput, verification nodes still have to duplicate the computation fully.
As the security of a decentralized system is directly related to the number of independent participants working to secure the network, Flow supports large numbers of participants with diverse technical, as well as financial commitments, leading to a system with low barriers to participate, while being costly to subvert.
Go-to-market: mainstream brands to drive mainstream adoption
Dapper Labs has been fantastic at winning over household brand names, first and foremost the NBA to develop the Top Shot collectibles experience as well as a triple A mobile game (16% of US americans surveyed are avid fans, 31% casual fans + large international fanbase). NBA Top Shot, which captures the nostalgia of trading cards and the thrill of sneaker trading, but within a digital universe, is exceeding all expectations in testing: over $1.2m of revenue with the first 500 players (out of the 12k+ prelaunch waitlist).
Beyond that major publicly announced partnerships include Warner Music Group, Ubisoft, Dr. Seuss Entertainment and UFC.
We thrilled to be part of the Flow community and are looking forward to seeing the infinite possibilities in truly user-owned content, digital collectibles/accessories as well as applications being realized over the coming years.
For more information on Flow visit onflow.org, Dapper Labs’ medium page and join the discord channel. To start playing NBA Top Shot, visit nbatopshot.com or sign up for beta access here.