The Graph vs. Self Hosting
Reading time: 3 min
The Graph’s decentralized network has been engineered and refined to create a robust indexing and querying experience—and it’s getting better every day thanks to thousands of contributors around the world.
The benefits of this decentralized protocol cannot be replicated by running a graph-node locally. The Graph Network is more reliable, more efficient, and less expensive.
Here is an analysis:
- Significantly lower monthly costs
- $0 infrastructure setup costs
- Superior uptime
- Access to hundreds of independent Indexers around the world
- 24/7 technical support by global community
No contracts. No monthly fees. Only pay for the queries you use—with an average cost-per-query of $40 per million queries (~$0.00004 per query). Queries are priced in USD and paid in GRT or credit card.
Query costs may vary; the quoted cost is the average at time of publication (March 2024).
| Cost Comparison | Self Hosted | The Graph Network |
|---|---|---|
| Monthly server cost* | $350 per month | $0 |
| Query costs | $0+ | $0 per month |
| Engineering time† | $400 per month | None, built into the network with globally distributed Indexers |
| Queries per month | Limited to infra capabilities | 100,000 (Free Plan) |
| Cost per query | $0 | $0‡ |
| Infrastructure | Centralized | Decentralized |
| Geographic redundancy | $750+ per additional node | Included |
| Uptime | Varies | 99.9%+ |
| Total Monthly Costs | $750+ | $0 |
| Cost Comparison | Self Hosted | The Graph Network |
|---|---|---|
| Monthly server cost* | $350 per month | $0 |
| Query costs | $500 per month | $120 per month |
| Engineering time† | $800 per month | None, built into the network with globally distributed Indexers |
| Queries per month | Limited to infra capabilities | ~3,000,000 |
| Cost per query | $0 | $0.00004 |
| Infrastructure | Centralized | Decentralized |
| Engineering expense | $200 per hour | Included |
| Geographic redundancy | $1,200 in total costs per additional node | Included |
| Uptime | Varies | 99.9%+ |
| Total Monthly Costs | $1,650+ | $120 |
| Cost Comparison | Self Hosted | The Graph Network |
|---|---|---|
| Monthly server cost* | $1100 per month, per node | $0 |
| Query costs | $4000 | $1,200 per month |
| Number of nodes needed | 10 | Not applicable |
| Engineering time† | $6,000 or more per month | None, built into the network with globally distributed Indexers |
| Queries per month | Limited to infra capabilities | ~30,000,000 |
| Cost per query | $0 | $0.00004 |
| Infrastructure | Centralized | Decentralized |
| Geographic redundancy | $1,200 in total costs per additional node | Included |
| Uptime | Varies | 99.9%+ |
| Total Monthly Costs | $11,000+ | $1,200 |
*including costs for backup: $50-$100 per month
†Engineering time based on $200 per hour assumption
‡Reflects cost for data consumer. Query fees are still paid to Indexers for Free Plan queries.
Estimated costs are only for Ethereum Mainnet subgraphs — costs are even higher when self hosting a graph-node on other networks. Some users may need to update their subgraph to a new version. Due to Ethereum gas fees, an update costs ~$50 at time of writing. Note that gas fees on are substantially lower than Ethereum mainnet.
Curating signal on a subgraph is an optional one-time, net-zero cost (e.g., $1k in signal can be curated on a subgraph, and later withdrawn—with potential to earn returns in the process).
Zero setup fees. Get started immediately with no setup or overhead costs. No hardware requirements. No outages due to centralized infrastructure, and more time to concentrate on your core product . No need for backup servers, troubleshooting, or expensive engineering resources.
The Graph’s decentralized network gives users access to geographic redundancy that does not exist when self-hosting a graph-node. Queries are served reliably thanks to 99.9%+ uptime, achieved by hundreds of independent Indexers securing the network globally.
Bottom line: The Graph Network is less expensive, easier to use, and produces superior results compared to running a graph-node locally.