Parameter Change Proposal: Increase number of active transcoders from 15 to 25

This post is a call for feedback from the community on a proposal to increase the number of active transcoders in the Livepeer protocol from 15 to 25 during round 1261, taking effect during round 1262, on approximately Wednesday, February 20.

Background

During Livepeer’s alpha release (Snowmelt Phase), the number of active transcoders on the network was set to be intentionally small at 10, allowing the network to be tested with a small group of actors before expanding. During round 1052, this limit was expanded by 5 in the first parameter update to the protocol to 15 active transcoders.

The intention is to scale well beyond this small number, and the upcoming Streamflow upgrade intends to enable hundreds to thousands of active nodes. However, in the meantime, there are a number of reasons to consider raising the limit to 25 active transcoders.

Proposed Benefits of the Change

  • Increase the decentralization of the network by encouraging greater participation from new actors.
  • The amount of stake delegated towards nodes in the active set has created quite a high bar for entry into the set, discouraging new, aspiring node operators, from considering a candidacy for Livepeer transcoder.
  • Since revealing statistics on the explorer of transcoder protocol interaction performance, the active set has been performing admirably missing very few reward calls. There should be room for more actors to achieve this level of performance.
  • The more nodes who are familiar with running the software and performing this role prior to Streamflow testing and rollout, the more participants in the ecosystem there will be to help QA, test, and prepare for the changes that Streamflow will bring.
  • The protocol has run stably and consistently for many hundreds of rounds, so the technical risk to introducing this small scope change is derisked relative to near the launch of the alpha.

Check out the existing transcoder dashboard.

Check out the existing transcoder social campaign threads here in the forum.

Potential Drawbacks of the Change

  • It will be 62% more expensive in terms of gas to call the initializeRound() txn, which will now require about 4,500,000 gas. This txn is invoked 1x/day across the entire protocol. Since this is approximately 50% of the current block gas limit, it is still believed this can be confirmed quickly, and in Streamflow the intention is to move to an implementation that doesn’t require this sort of linear overhead in the initializeRound txn as more transcoders are added anyway.
  • Bond, unbond, rebond transactions will increase approximately 8% in gas usage/cost. This affects every user, but since these actions are infrequent and low cost to begin with, hopefully the increase in decentralization and benefits listed above. As an example, a bond transaction may increase from approximately 278,000 gas to 300,100 gas, or essentially by $0.01 in USD terms at 5 gwei gas prices.
  • If more stake activates on the ten newly added nodes, then existing transcoders may see their reward call token portions diminished as a result of greater decentralization of the newly generated tokens. For any individual actor, you can still expect to receive rewards in direct proportion to your stake, however if the stake is spread out further than the reward cuts for transcoders may result in less overall reward tokens. Transcoders can adjust for this by increasing reward cut if they choose. With a limited set of transcoders, each has a large share of a small pie, however the belief is that by increasing general participation, the ecosystem will benefit from more network incentive alignment leading to faster experimentation and more growth, increasing the size of the pie for all.
  • DApps and Apps that reference Livepeer and have hardcoded assumptions around the 15 transcoder limit may need to update copy or UX. If they have pulled the limit from the protocol itself, they would adapt automatically, but it may take them time to discover the protocol update and make manual changes if necessary. This likely would not create breaking experiences, but it’s possible. We will proactively reach out to give everyone notice if this proposal leads to a parameter change.

Open discussion

Have any concerns of questions about this parameter upgrade? Share them here. As part of a shift from core team parameter setting to decentralized governance all input and discussion is welcome in an open way before executing any updates to the protocol.

3 Likes

This proposal feels like a a good trade-off in the short term. Transcoders like @0x88(Transcoder Campaign: 0x88 Hyper Transcoder) and @dillchen(Transcoder Campaign: Commonwealth Transcoder) will be included in the active set, and the lower activation threshold is good for more decentralization of the network.

I expect the newly opened slots to be quickly filled up due to the increasing awareness of staking opportunities in general. Due to the large % of increase in gas cost for initializeRound(), I don’t expect the numbers to increase fast enough to capture all demand for new nodes. So it is important for a protocol mechanism update to happen in order to allow hundreds to thousands of nodes on the network.

Looking at the future of the network, especially around the launch of Streamflow, I expect there to be some turnover in the active set as the responsibility of a Livepeer node will be more about providing transcoding capacity to the network. I’m looking forward to that day, when the Livepeer network can provide transcoding service at 1/10 of the price comparing to other cloud transcoding services.

3 Likes

As operator of the Chainflow Livepeer Transcoder, I support this proposal :+1:

The increased gas price, potentially reduced rewards and stake feel like worthwhile investments, in order to continue decentralizing the network :muscle:

1 Like

Wowwww… Thanks you… loveeeeeeeee this proposal

1 Like

I’m looking into Aragon to build a DAO-backed transcoder, so I am all for it!

2 Likes

I support this proposal ,we need more active transcoders join it

As a quick update, due to the ice age on Ethereum, block times have increased to 20 seconds, and therefore round 1261 is unlikely to occur until Sunday 2/24. This is a couple days later than originally proposed.

1 Like

Out of interest, what was the process to agree this change?

  1. Receive many requests from the community for increase in transcoder set size.

  2. Attempt to understand and reason about the economic incentive effects of this change, the technical effects, the implementation details, and risks associated with the change.

  3. Propose the change publicly via this proposal, and call for community feedback.

  4. Socialize the proposal on social channels - Discord, Forum, Twitter.

  5. Review feedback and evaluate concerns.

  • if there had been any impression of concern, then consider more formal methods for signaling such as transcoder/delegator poll or carbon vote type methods.
  • had there still been significant source of contention, then hold off on any change until it could be enacted by means such as fork or decentralized governance parameter enactment in the future.
  1. As there has been no voiced opposition, and much voiced support, the change can move forward in round 1262.
2 Likes

The execution of this parameter change was completed as of block 7266263 (Feb-25-2019 02:32:52 PM +UTC) during round 1261. At this time, the maximum number of registered transcoders in the network is 30 and the maximum number of active transcoders during a round in the network is 25. Additional transcoders can now be registered up to the new cap of 30 (the previous cap was 20). When round 1262 is initialized, the top 25 transcoders with the most stake will be locked in as the active set for the round.

3 Likes