Evaluate Running a Passport ETH Archive Node

[Jkd-eth]

User Story:

As the Passport team,
I want to have unlimited and cost-effective access to Ethereum transaction data,
So that we can efficiently run analyses and query the blockchain without incurring high third-party costs.

Acceptance Critiria

GIVEN the need for detailed blockchain data analysis,
WHEN evaluating different options for hosting an Ethereum archive node,
THEN we should decide on the most cost-effective and efficient setup for Passport.

Exploration Goals

1. Evaluate Feasibility: Explore running an Ethereum archive node using Reth and Trueblocks, including the option of using dedicated hardware.
2. Cost Analysis: Determine the monthly operating cost for each option, both cloud-based (e.g., AWS) and local hardware setups.
3. Comparison: Compare Reth, Trueblocks, and other potential solutions, focusing on functionality, cost, and ease of integration.
4. Decision Making: Provide a go/no-go recommendation based on findings.
5. Next Steps: Outline subsequent stories and tasks if the decision is to proceed with running our own node.

Product & Design Links:

Tech Details:

• Previous cost estimation of $500/month for AWS setup: Cost Estimate
• Hardware specifications and requirements for local setup
• Reth and Trueblocks installation and configuration details

Open Questions:

• Can this setup support multiple blockchains, or is it limited to Ethereum?
• How does this impact our current reliance on Alchemy, and will it reduce our associated costs?

Notes/Assumptions:

• Assumption: We have access to the required hardware or cloud infrastructure for testing.
• Note: The exploration should not exceed 3 days to keep costs and time investment reasonable.

[erichfi]

Let's see if we still need a centralized transaction data cache in case we run an Archive Node: #2829

[tim-schultz]

Ethereum Node Requirements

Mainnet Nodes

RETH Mainnet

• Disk: 2.2TB+ (TLC NVMe recommended)
• RAM: 8GB+
• Note: NVMe disk crucial for I/O performance

GETH Mainnet

• RAM: 16GB+ recommended
• Disk: 12TB+ for full archive node

Layer 2 Solutions

Optimism / OP Stack

• RAM: 16GB+
• Disk: 2TB SSD (NVMe recommended)

Arbitrum

• RAM: 16GB
• CPU: 4 cores (single-core performance important)
• Storage (as of April 2024):
  • Arbitrum One: 560GB (pruned), growing ~200GB/month
  • Arbitrum Nova: 400GB (pruned), growing ~1.6TB/month
• Note: NVMe SSD recommended for both

Polygon

Component	Minimum	Recommended
CPU	4 cores	16 cores
RAM	32GB	64GB
Storage	2.5TB	5TB
Bandwidth	100 Mbps+	1 Gbps
AWS instance	c5.4xlarge	m5d.4xlarge

AWS Configurations for Polygon Nodes

Basic Node (single to double-digit RPC requests/s)

• Compute: m7g.4xlarge (16 vCPU, 64GB RAM)
• Storage: 7000 GB EBS GP3, 16000 IOPS, 1000 MBps throughput

High-Performance Node (hundreds of RPC requests/s)

• Compute: im4gn.4xlarge (16 vCPU, 64GB RAM)
• Storage: 7500 GB instance storage, 50 GB EBS gp3 root volume

Maximum Performance Node (up to 1000 RPC requests/s)

• Compute: m7g.4xlarge (16 vCPU, 64GB RAM)
• Storage: 7000 GB EBS IO2 with 16000 IOPS

Ethereum Node Testing and Deployment Summary

Local Testing

Hardware Specifications

• RAM: 32GB
• Storage: 4TB SSD
• Network: ranged between 5 - 15 Mb/s upload and download speeds

Equipment

• Base System: Amazon Link
• Upgraded Disk: Amazon Link

Node Synchronization

• Used RETH from Merkle.io snapshots
• 1.3TB file downloaded and extracted overnight (approximately 12 hours)
• Node is keeping up with the network tip, including pending blocks

Performance Test

• Downloaded all USDC transfer events to parquet files (Block 0 to present)
• Command: cryo erc20_transfers --contract 0xa0b86991c6218b36c1d19d4a2e9eb0ce3606eb48 --rpc localhost:8545
• Duration: 1 hour and 10 minutes
• Additional data types available: Cryo Datasets

Data Query Performance

• Specific address transfer events retrieved in 6 seconds
• Potential for joining data on a large number of addresses
• Example query: Gist Link
• Could also build a well indexed DB for pulled data

Running Additional Nodes

• L2 nodes need to run in parallel with the mainnet node
• Additional storage required for each L2 node
• Lack of good L2 snapshot sources noted

ISP Requirements

• Unlimited bandwidth is necessary

Pricing Considerations

Latitude (Suggested by RETH)

• Model: rs4.metal.large
• Price: $2.20/hr or $1606/month
• Storage: $0.64/TB
• Capability: Could theoretically run all currently used nodes
• Note: Can be turned on/off as needed, but requires sync time

AWS

• Model: m6gd.16xlarge
• Price: $2.8928/hr or $2108/month
• Additional EBS storage costs apply
• Potential for optimization using powerful processor with cheaper EBS storage

Kammerdeiner

• Pricing: TBD

Individual Network Pricing

• If certain networks are a priority, pricing can be obtained for individual networks

Other Considerations

• Time-consuming to ensure and maintain node functionality
• Potential efficiency in bulk data pulling from current Alchemy account on a regular basis
• Further discussion needed on feasibility of using these tools for analysis

Saving on Costs

• Cryo/TrueBlocks are awesome tools for data extraction from RPC nodes
• We could consider bulk pulling data from our current Alchemy account and running analysis on our data(postgres or parquet) if this would work for the data team