One of the most critical things for the Livepeer community to address in the coming months is to ensure that there’s better alignment between those doing useful video transcoding work on the network, and the rewards that are earned through protocol token rewards and fees. The fees flowing through the network are rising, but are still not high enough to eclipse the value of the daily LPT rewards that the protocol issues, and therefore many node operators do not yet view it as worth their time and effort to compete for high performance video transcoding work, which is what the network needs in order to be highly useful and cost effective to video developers at scale.
While ultimately the community can consider incentive changes within the protocol itself to better align these interests, in the short term there are a number of tools and mechanisms at our disposal to affect action here. The first two, which are focused on adding visibility, are a stream testing tool and accompanying orchestrator leaderboard that showcase each orchestrator’s performance in the explorer.
The way that the leaderboard will work is roughly as follows:
- Multiple times per day the test harness will connect to each orchestrator, send it some number of concurrent streams (1–3 initially, then more), and capture various statistics about the performance. This will be paid work through the network.
- This harness will run in three different regions: US (Chicago), Europe (Frankfurt), Southeast Asia (Singapore), and collect statistics for each.
- For each region, a leaderboard will show a transcoding success rate and a combined score over the past 24 hours that takes into account latency, success, error types, and availability
The resulting output should increase visibility into which nodes are actually contributing transcoding value to the network, and are equipped to support the network as it scales in usage. On the other side of the coin, it should also showcase those that have been collecting large amounts of inflationary LPT, without contributing valuable compute resources to perform video encoding.
What does this mean for orchestrators?
Orchestrator operators should strive to achieve high scores on the leaderboard. This means running with enough capacity to encode streams in sub-realtime, with good bandwidth, and always on availability. Operators should check out the docs and FAQs to get informed on how to best run their infrastructure.
What does this mean for token holding delegators?
Those who have staked on the network should pay attention to the leaderboard, and consider moving their stake towards those who are performing well. This will serve two purposes:
- It will incentivize node operators to run their infrastructure in a way that benefits the network, because the extra stake they attract means they can earn a larger amount of LPT through their reward cut.
- It will provide more security and route more work towards nodes that are equipped to handle it, making the Livepeer network useful and cost effective for video developers looking to pay fees into the network. Ultimately, it contributes to making the network succeed long term.
How to reward high performing nodes?
As previously stated, it is important that the network rewards those that are contributing valuable work. These rewards can come in the form of fees, LPT rewards, and external incentives. Fees are the ideal long term incentive, but at the current moment it’s important for the community to use the tools at its disposal to ensure rewards and external incentives drive the right type of participation.
Inflationary LPT rewards, delegation, and inflation funding are concepts already baked deeply into the Livepeer protocol that can be useful here. In particular, it is part of a delegator’s responsibility, and in their long term interest, to delegate stake towards those node operators that are helping the network achieve its potential. While delegating purely on reward cut to maximize short term LPT gains may look attractive, if the value of LPT falls because the network is not useful, this is short sighted and against one’s self interest. Instead, token holders should find a balance between delegating to nodes who are visibly helping the network, and sharing back appropriate rewards and fees to make the long term calculation optimized in the delegator’s favor.
Visibility into performance through the leaderboard should help with this dramatically, and if large amounts of stake begin to swing towards those nodes who are performing well, they’ll be rewarded with increased LPT and be able to invest more deeply in supporting the network by adding more GPU encoding capacity. While one may think initially that this dynamic would lead to centralization of stake on a few high performers, it is actually likely that this dynamic is more open and accessible for low-stake newcomers to immediately make an impact, break in, and accrue stake quickly. The default transcoder view on the explorer is ranked by stake, focusing attention only on those who have ossified their position at the top of the stake list, but this performance leaderboard is ranked by performance score, and therefore even a node operator with 1 LPT staked could be shown alongside or above the highest staked nodes. If the above dynamic holds true, then the community should shift stake towards this node, for a larger portion of the fee pie shared back, and the result should be that pretty quickly stake shifts towards a number of newly competitive, high performing nodes that can form the basis of a reliable, global, transcode network.
External Incentives and Updated Protocol Economics?
While the above thesis about increased visibility resulting in the shifting of stake and inflationary rewards still needs to be tested, it is also worth thinking about other mechanisms that can be used to route rewards towards those performing valuable work. One such idea that is external to the protocol follows the model of incentivized testnets that other networks have put on. These programs could be funded by community nodes, grant pools, community DAOs, or even inflation funding. They can take the soft rules behind the reliance on token holder staking action to route rewards, and make them more concrete, tying evidence of solid network contribution to hard rewards.
Likewise, the community should also get thinking about how to go from an externally administered reward program like the one just mentioned, to something that is on-chain administered. Proving real, non-gamed, valuable contribution on chain remains a tricky and possibly-impossible-to-solve research problem, but there is a creative design space for a middle ground here. Imagine a protocol update that allocates a reward pool to be distributed for the purposes of continuing to bootstrap valuable GPU mining supply side work, and the protocol’s governance mechanism being used to put in place the rules for the distribution and release of these rewards.
The usage of the Livepeer network is growing as more and more applications build on and onboard onto the platform. Visibility into performance of the node operators has been subpar to date, but now that it can be put front and center via the new performance leaderboard, the community can be catalyzed into action to ensure that the incentives are there for valuable work to be performed. This is just the beginning of this track of work, as a single leaderboard based on a single characteristic isn’t possibly enough to capture all the forms of contribution to the network — but it’s a starting point that we should collectively engage around, and use to shape the future of Livepeer network’s supply side.