Nostr has rapidly emerged as a groundbreaking force in the decentralized web, offering a censorship-resistant, user-owned alternative to traditional social media platforms. Unlike centralized giants such as Twitter or Facebook, Nostr operates on a unique relay-based architecture that separates user identity from server infrastructure, enabling unprecedented levels of decentralization and resilience. In just two years since its 2022 launch, the network has grown to over 4 million users and 60 million posts — sparking intense interest from technologists, privacy advocates, and researchers alike.
This article presents a comprehensive analysis of the Nostr ecosystem based on a large-scale measurement study conducted between July 1 and December 31, 2023. Drawing from a dataset of 17.8 million posts, 1.5 million public keys (users), and 712 active relays, we explore how Nostr achieves decentralization, assess its strengths in post availability, and identify critical challenges related to relay sustainability and system overhead.
Core Architecture of Nostr
At its foundation, Nostr (Notes and Other Stuff Transmitted by Relays) is built on a minimal yet powerful set of open protocols. The system consists of two primary components: clients and relays.
- Clients: These are user-facing applications — available on web, iOS, Android, and desktop — that allow individuals to create, sign, publish, and retrieve messages.
- Relays: These function as stateless storage servers that receive and distribute signed messages (called events) without verifying content or managing identities.
Unlike federated platforms like Mastodon where servers communicate with each other (via ActivityPub), Nostr relays do not interact. Instead, users connect directly to multiple relays through their clients, publishing the same message across several nodes simultaneously. This decoupling of identity from infrastructure ensures that no single point of failure can silence a user.
Each user is identified solely by a cryptographic public key, with all messages digitally signed using their private key. This model eliminates reliance on usernames, passwords, or centralized account systems — granting true ownership and portability.
👉 Discover how decentralized networks are reshaping digital identity and communication
Measuring Decentralization Across Relays
One of Nostr’s most compelling promises is its ability to resist centralization — a persistent flaw in earlier decentralized networks like the Fediverse.
Our study analyzed the distribution of posts and users across 712 relays. We found that:
- 93% of posts exist on multiple relays, with an average replication across 34.6 relays per post.
- The top relay hosts 73% of all posts — but due to widespread replication, this does not indicate centralization.
- 178 relays (25%) host more than 5% of total posts, demonstrating broad content distribution.
- Relays are spread across 44 countries and 151 autonomous systems (ASes).
- No single country hosts more than 25% of relays; 64% of ASes run only one relay.
These findings reveal a highly decentralized infrastructure — significantly surpassing the concentration seen in Mastodon instances during prior studies. Even when major countries like the US host large volumes of data, the multi-relay replication model ensures redundancy and resilience.
However, decentralization alone doesn’t guarantee reliability — which leads us to a critical challenge.
Relay Availability and Financial Sustainability
Despite strong structural decentralization, relay uptime remains inconsistent.
- 20% of relays experience downtime for over 40% of the time.
- 132 relays were completely inactive ("dead") by the end of our monitoring period.
- Larger relays (hosting >1M posts) show better outage recovery — only 1% of outages exceed 1000 minutes vs. 88% for smaller ones.
A key factor behind instability appears to be financial sustainability, especially for free-to-use relays.
Using Bitcoin-powered "zaps" (micro-tips), users can support relay operators. However, our analysis shows:
- 64% of relays receive no zaps at all.
- Among those that do, the 90th percentile earns only ~150,000 satoshis (~$67 USD) over six months.
- Even with admission fees (on paid relays), 95% of free relays cannot cover operational costs.
This suggests that most relays are maintained by enthusiasts rather than sustainable operations — posing long-term risks to network health.
👉 Learn how blockchain-based incentives are transforming digital infrastructure
Post Availability vs. System Overhead
Nostr’s design excels in ensuring message availability even during partial network failures.
We simulated catastrophic scenarios by removing the top relays and ASes:
- Even after removing the top 50 relays by user count, over 90% of posts remain accessible.
- After removing the top 10 ASes, over 80% of posts are still available — far exceeding Fediverse resilience.
This robustness stems from aggressive post replication. However, it comes at a steep cost:
- Total post replications: 616 million for just 17.8 million unique posts.
- Average replication: 34.6 relays per post.
- Estimated wasted retrieval traffic: 144 TiB over six months.
- 98.2% of client downloads are redundant, as users fetch the same post from multiple relays.
This inefficiency creates unnecessary strain on both bandwidth-constrained clients and relay operators paying for data transfer.
Design Innovations to Reduce Overhead
To address these inefficiencies without compromising availability, we propose two protocol-level improvements.
Strategic Replication Reduction
Instead of broadcasting every post to all configured relays, clients can adopt smarter strategies:
- Random Selection: Limiting replication to 20 randomly chosen relays reduces redundant copies by 380 million while maintaining >97% availability.
- AS-Based Selection: Prioritizing relays across diverse autonomous systems improves fault tolerance — achieving near-original availability with only 15 relays.
- Follower-Based Selection: Targeting relays used by followers increases delivery likelihood but risks centralization on popular nodes.
Our evaluation shows that reducing maximum replications to 10–20 relays dramatically cuts storage overhead with minimal impact on availability.
Optimistic Retrieval
To reduce redundant downloads, we introduce optimistic retrieval:
- Users publish a list of “recommended relays” (using NIP-65).
- Followers retrieve regular ("low-priority") posts only from these recommended relays.
- High-priority posts (e.g., urgent updates) are still fetched from all connected relays via a new
prioritytag.
Results show:
- With just 5 recommended relays, clients achieve 99.4% post completeness.
- Wasted retrieval drops from 98.2% to 11.7%.
- Bandwidth savings exceed 88%, reducing load on both clients and servers.
This approach balances efficiency with reliability — ensuring critical content reaches all users while optimizing routine communication.
Frequently Asked Questions (FAQ)
What makes Nostr different from Mastodon or other Fediverse platforms?
Nostr eliminates server-to-server federation. Users publish directly to multiple independent relays using cryptographic keys, making accounts portable and resistant to censorship. Unlike Mastodon, where your identity depends on an instance admin, Nostr gives full control to the user.
Can I run my own Nostr relay?
Yes — anyone can set up a relay with minimal technical requirements. Open-source implementations exist in various programming languages, and many operators run them on low-cost VPS instances. However, long-term sustainability requires either community support or monetization through zaps or access fees.
How does Nostr handle spam or abuse?
Nostr lacks built-in moderation tools. Instead, clients implement filtering based on blocklists, trust networks, or AI models. Some services offer shared moderation databases, but enforcement remains decentralized — placing responsibility on end users.
Why are posts replicated so many times?
Replication ensures high availability and censorship resistance. If one relay goes offline or blocks content, copies remain available elsewhere. However, current defaults lead to excessive redundancy — which new optimization strategies aim to reduce.
Does Nostr use blockchain?
Not natively. While Nostr supports Bitcoin-powered zaps via the Lightning Network, the core protocol doesn’t rely on any blockchain. It uses public-key cryptography and WebSockets for secure messaging — keeping the system lightweight and fast.
How can developers contribute to Nostr?
Developers can build clients, improve relay software, propose new NIPs (Nostr Implementation Possibilities), or create tooling for analytics and moderation. The ecosystem thrives on open collaboration — with active discussions in public Nostr channels and GitHub repositories.
👉 Explore developer tools powering next-generation decentralized applications
Conclusion
Nostr represents a significant leap forward in decentralized social networking. Its relay-based architecture achieves superior decentralization compared to existing platforms, with impressive resilience against outages and censorship. However, challenges remain — particularly around relay sustainability and protocol inefficiencies caused by excessive replication.
Through data-driven analysis, we’ve demonstrated that strategic optimizations — such as limiting replication scope and adopting optimistic retrieval — can drastically reduce system overhead without sacrificing availability.
As Nostr continues to evolve, these insights provide a roadmap for building a more efficient, scalable, and sustainable decentralized web. For developers, researchers, and users committed to digital freedom, Nostr offers not just an alternative — but a blueprint for the future of open communication.
Core Keywords: Nostr ecosystem, decentralized social network, relay availability, post replication, censorship resistance, blockchain incentives, Nostr clients