Network Metrics Reference

Every Livepeer Network metric: where to find it, what it means, what surface it comes from, and how to read it. The full reference.

Primary network metrics are emitted by the following surfaces:

Read at: explorer.livepeer.org

Cadence: Live; charts update on each round

Auth: None for read access

Visit Explorer

Read at: gateway.thegraph.com/api/{API_KEY}/subgraphs/id/FE63YgkzcpVocxdCEyEYbvjYqEf2kb1A6daMYRxmejYC

Cadence: Indexed per Arbitrum block (~250ms)

Auth: The Graph API key required

Open Subgraph

Read at: leaderboard-serverless.vercel.app/api/aggregated_stats

Cadence: Recomputed on test-job runs (~hourly)

Auth: None

Open API

Read at: nyc.livepeer.com/orchestratorStats

Cadence: Updated by orchestrators advertising prices

Auth: None

Open Endpoint

Read at: <AI_METRICS_SERVER_URL>/api/pipelines

Cadence: Per orchestrator capability advertisement

Auth: None

View Reference

Round Status

The protocol advances in rounds. A round is roughly 21 hours on Arbitrum One; the precise length is the roundLength parameter measured in L1 blocks. Round state controls when orchestrators can call reward, when stake changes take effect, and when daily aggregates close.

Economic Metrics

LPT stake metrics measure how much LPT is bonded to orchestrators, what proportion of supply that represents, and how stake is distributed across the network. These metrics determine the active set, drive inflation, and gate every governance vote.

Decentralisation Metrics

The active set is the bounded list of orchestrators eligible to receive video work each round. Membership is determined by stake. These metrics measure the size, composition, and turnover of the set.

Staking Metrics

Each orchestrator has its own stake position. These metrics describe how much LPT is bonded to a specific orchestrator and how that stake is split between the orchestrator’s own bond and delegators.

Commission Parameters

Reward cut and fee cut are the orchestrator’s commission rates: how much of the round’s LPT inflation rewards and ETH fees the orchestrator keeps before splitting the rest to delegators by stake share.

Reward Pools

Each round, every active orchestrator can call BondingManager.reward() to mint that round’s share of inflation into a Pool. Pools record what was minted, who minted it, and how it gets split.

Activity Metrics

Fee volume is the ETH paid for jobs settled through winning tickets, with USD-equivalent computed at settlement time. The same fee flow generates per-orchestrator metrics, per-gateway metrics, and the network-wide aggregate.

Usage Estimates

Usage is converted from fee volume into estimated minutes of compute delivered. The conversion methodology was last updated 21 August 2023 (see ).

Performance Leaderboard

The leaderboard is populated by a metrics server (leaderboard-serverless.vercel.app for transcoding, separate URL for AI) that runs reference test jobs against every active orchestrator from each region. Scores reflect how each orchestrator answers those test jobs.

Regions

The leaderboard runs from a fixed set of measurement regions, each typed by the workload it tests.

AI Capabilities

AI pipeline capabilities are advertised by orchestrators and aggregated by the AI metrics server. These metrics describe what AI work is available on the network and where.

Gateway Columns

Gateways prepay for jobs by funding a deposit and a reserve in the TicketBroker contract. These metrics show the prepaid funds and how recently the gateway has been active.

Treasury and Governance

The treasury is a separate on-chain pool of LPT funded by a percentage of each round’s inflation. Treasury proposals authorise spending against it.

Governance Voting

Stake-weighted voting on protocol parameter polls and treasury proposals. Vote weight is the voter’s totalStake at the snapshot block.

Transactions Feed

Every state-changing protocol call is indexed as an event on the subgraph and shown on the Explorer transactions feed.

Operator Metrics

These metrics live on the operator’s own go-livepeer node, not on the Explorer or subgraph. Each operator exposes them through the go-livepeer Prometheus endpoint and CLI; no central party aggregates them.

Subgraph Metadata

Every subgraph response includes indexer metadata. Use this when computing freshness or when reproducing a query at a specific block.

Endpoints

For people writing scripts or dashboards, the canonical endpoints in one block:

Off-Chain Blind Spots

Three classes of data exist but are not exposed publicly. Anyone building dashboards or doing research should know what is missing.

• Per-segment job content. Individual transcoded segments and AI inference outputs are exchanged off-chain between gateway and orchestrator. They are not retained, not indexed, and not visible after delivery.
• Non-winning probabilistic tickets. Most tickets do not win and are never submitted on-chain. The off-chain ticket exchange is private to the gateway-orchestrator pair.
• Per-session price agreement. Specific orchestrator selection decisions and per-job price agreements happen during the gateway-to-orchestrator handshake. Aggregate pricing is visible (pricePerPixel, capability advertisement); session-level decisions are not.

These omissions are by design. The marketplace runs at a cadence the chain cannot match; what stays off-chain is high-frequency ephemeral coordination with no audit value once the work is delivered.

Related Pages

The five surfaces these metrics come from, end-to-end.

The Network topology these metrics describe.

What fee volume, settlement, and ticket redemption represent.

Contract addresses backing every on-chain metric on this page.