Skip to content

Reduce Heima parachain block time 6s → 2s via elastic scaling (3 coretime cores) #4041

Description

@hanwencheng

Summary

Drive Heima's parachain block time from ~6s to ~2s using elastic scaling on 3 coretime cores, giving latency-sensitive EVM applications Base-class responsiveness while keeping Heima's much cheaper gas.

Why

Per-op gas on Heima is already excellent — the historic heavy cost (the pure-Solidity P-256 passkey verify, ~1M gas) is gone now that RIP-7212 is live (runtime 9261): K11 verify is ~28k gas. The only axis where an L2 like Base still wins is latency: Base's sequencer produces ~2s blocks vs Heima's ~6s. For interactive flows a user-facing write costs ≈ one block of wait, so block time is the perceived latency.

Current state (please confirm)

Heima appears to run ~6s blocks, which suggests async backing is already enabled (the 12s→6s win is banked). Confirming this is step 1.

Mechanism

Heima is a Polkadot parachain (para_id 2013). The relay already ships async backing + elastic scaling + agile coretime (Polkadot SDK 2509, Oct 2025), so the capability is present — the remaining work is Heima-side (runtime + collator) plus acquiring cores.

A parachain block is capped at 2s of execution; the relay slot is 6s. Elastic scaling lets one parachain occupy multiple cores in parallel, one block per core per slot:

Cores Blocks / 6s relay slot Block time
1 (today) 1 ~6s
2 2 ~3s
3 3 ~2s

This also raises aggregate blockspace ~3× — throughput and latency both improve, it is not a trade.

Proposed change (runtime + collator)

Knob Today Target
MILLISECS_PER_BLOCK / SLOT_DURATION 6000 2000
BLOCK_PROCESSING_VELOCITY 1 3
UNINCLUDED_SEGMENT_CAPACITY ~2–3 raise to match velocity (≈ velocity · relay/para + 1)
RELAY_CHAIN_SLOT_DURATION_MILLIS 6000 6000 (fixed by relay)
Collator binary lookahead collator slot-based collator (required for >1 core)
Coretime (Agile Coretime) 1 core 3 cores on para 2013

The permissioned (Litentry-run) collator set qualifies for the elastic-scaling MVP (trusted-collator phase — no candidate-receipt change). Validate on Heima Paseo (para_id 2013 testnet) before mainnet.

Diagnostics requested

  1. Current mainnet MILLISECS_PER_BLOCK.
  2. Polkadot SDK / runtime version (slot-based collator needs ≥ the 2509 line).
  3. Collator type (lookahead vs slot-based).
  4. Cores currently assigned to para 2013.
  5. Budget appetite for 3 cores.

Costs / risks

  • ~3× recurring coretime spend (3 cores on Polkadot).
  • Collator hardware: 2s blocks at ~100% duty cycle need collators that import + author a 2s-weight block in < 2s wall-clock.
  • Coordinated runtime + collator upgrade, tested on Paseo first.

References

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions