SONDA

Solana Observatory for Network Decentralization Analysis

Network analytics and tactical infrastructure intelligence for Solana validator operators.

Live dashboard: sonda.network Public data: data.sonda.network Project Twitter: @SondaNetwork

Note

SONDA is built and maintained by Solya Validator as a public good. The pipeline runs continuously on production, updating every 60 seconds for mainnet and alpenglow-community, every 5 minutes for testnet and devnet. Phase 1 dashboard shipped as part of Colosseum Frontier 2026. Phase 2 pages (per-validator, per-datacenter) are in active development.

Why SONDA?

SONDA is built for a specific question: how to make tactical decisions about Solana validator infrastructure. Operators face concrete trade-offs every week, and the data needed to answer them is scattered across half a dozen tools.

Tactical placement for operators. Where should the next validator node go? Which cities and ASNs already concentrate too much stake? Which datacenters have a track record of stability versus recurring delinquency? SONDA aggregates the data you would otherwise collect by hand.

Per-validator and per-datacenter history. Validators move datacenters. Operators switch to backup nodes. ASNs grow and shrink. SONDA tracks all of this over time, with location history going back to epoch 196 (2021). Per-validator and per-datacenter pages are rolling out in Phase 2.

Verified data foundation. Every IP is checked against four independent geolocation providers (DB-IP, IPInfo, GeoJS, ip-api), with DoubleZero device locations used as protocol-verified ground truth. Without this foundation, every map and metric on top of it would inherit the errors of whichever single source was chosen.

Multi-cluster coverage. Mainnet, testnet, devnet, and the experimental alpenglow-community cluster where Anza is testing the Alpenglow consensus algorithm. Each cluster looks very different. SONDA covers all four with the same pipeline.

Decentralization metrics as foundation, not headline. Nakamoto, HHI, Gini, and Shannon entropy across four dimensions (country, ASN, city, validator) are computed every cycle and published as part of the data. They support the tactical pages rather than being the primary product.

Who is this for?

Audience	What SONDA gives them
Validator operators (primary)	Tactical decisions on placement, datacenter and ASN selection, backup-node patterns, infrastructure timing
Stake pool operators	Due diligence on validator candidates, datacenter and provider risk visibility
Researchers and journalists	Measurable decentralization data across four dimensions, full historical context
SOL stakers	Where stake actually concentrates, which providers and regions dominate

What SONDA Analyzes

Network Nodes (5,000+ per scan on mainnet)

Category	Description
Validators	Active, hidden (no gossip), inactive (delinquent)
RPC Nodes	Public RPC endpoints in gossip
DoubleZero	DZ devices, connected validators, multicast groups
BAM / Jito	Block engines, shred receivers, NTP servers, BAM nodes
Harmonic	Auction engines, TPU relayers, bundles
Other	Entrypoints, co-hosted nodes, backup nodes

Geolocation: 4-source cross-verification

DB-IP ───── IPInfo ───── GeoJS ───── ip-api (used only for discrepancies)
                                       │
                       DoubleZero ─────┘  (protocol-verified ground truth)

Each IP receives a confidence score (high, medium, low) based on source agreement. Discrepancies are preserved with full alternatives for audit. Geo overrides from DZ devices automatically correct mislocated carrier IPs.

Decentralization Metrics

Metric	Dimensions	What it reveals
Nakamoto Coefficient	Country, ASN, City, Validator	Minimum entities to control 33% of stake
Superminority	Country, ASN, Validator (at 33 / 50 / 66%)	Geographic and organizational concentration thresholds
HHI	Country, ASN, Validator	Market concentration (competitive vs monopolistic)
Gini Coefficient	Validators	Stake inequality distribution
Shannon Entropy	Country, ASN, Validator	Diversity and evenness of distribution

Infrastructure Intelligence

Source	Data
DoubleZero	Device locations, validator connections, multicast groups
BAM	Node topology, validator mapping, IBRL performance scores, stake %
Rakurai	MEV-optimized validator detection, geo distribution
Trillium	Client types, vote latency, slot duration, SFDP status
Endpoints	90+ infrastructure endpoints with per-service reachability checks
Alpenglow	BLS pubkey adoption, feature activation status, genesis hash, rollback detection

Repository Layout

sonda/
├── analyzer/                       Core network analysis pipeline
│   ├── solana_analyzer.py          Real-time snapshot collector (2,800 lines)
│   ├── solana_history.py           Historical geo data collector (1,250 lines)
│   ├── endpoints.yaml              Infrastructure endpoint configuration
│   ├── geo_overrides.yaml          DZ-verified location overrides
│   └── README.md                   Detailed usage and reference
│
├── automation/                     Orchestration and pipeline glue
│   ├── run_sonda.py                Multi-cluster orchestrator (systemd entry point)
│   ├── run_once.py                 Single-cluster pipeline for testing
│   ├── split_snapshot.py           Splits 5MB snapshot into per-role files
│   ├── split_history.py            Splits history into per-validator files
│   ├── r2_upload.py                Cloudflare R2 upload (multi-mode)
│   ├── timeseries.py               SQLite state tracking, event detection
│   ├── telegram.py                 Public and debug bot notifications
│   ├── config.example.yaml         Configuration template
│   └── README.md                   Configuration and operation guide
│
├── systemd/                        Service definitions
│   ├── sonda.service               systemd unit (multi-worker orchestrator)
│   ├── sonda-logrotate             Log rotation config
│   └── README.md                   Installation and troubleshooting
│
├── .githooks/pre-commit            Blocks accidental secret commits
├── .gitignore                      Protects config.yaml with real keys
├── LICENSE                         MIT
└── README.md                       This file

sonda_data/ (snapshots, SQLite, history, R2 staging) lives separately at /home/solya/sonda_data/ on the production server. Not part of this repository.

Architecture

Real-time pipeline

run_sonda.py (systemd)
    │
    ├── Worker [mainnet-beta]            60s interval
    ├── Worker [testnet]                 300s interval
    ├── Worker [devnet]                  300s interval
    └── Worker [alpenglow-community]     60s interval
            │
            ▼
        Each worker cycle:
        ┌─────────────────────────────────────────┐
        │  1. analyzer/solana_analyzer.py         │
        │     ├── Gossip + validators (Solana CLI)│
        │     ├── Genesis hash, features          │
        │     ├── BLS pubkeys (Alpenglow only)    │
        │     ├── External APIs (BAM, DZ, etc.)   │
        │     ├── 4-source geolocation engine     │
        │     └── Write snapshot JSON (~5MB)      │
        │                                          │
        │  2. automation/split_snapshot.py        │
        │     └── 4 role-specific JSON files      │
        │                                          │
        │  3. automation/r2_upload.py             │
        │     └── Push to data.sonda.network      │
        │                                          │
        │  4. automation/timeseries.py            │
        │     ├── Diff against previous state     │
        │     ├── Record node_changes, ip_changes │
        │     ├── Detect cluster rollback         │
        │     └── Record epoch snapshot on wrap   │
        │                                          │
        │  5. automation/telegram.py              │
        │     └── Send events to bot channels     │
        └─────────────────────────────────────────┘

Cluster rollback detection

The alpenglow-community cluster experiences regenesis events approximately weekly and snapshot-based restarts more frequently. SONDA's timeseries layer distinguishes three rollback types using genesis_hash, epoch, and slot:

Scenario	Signal	Response
Regenesis	genesis_hash changed	Skip all comparisons, warn
Epoch rollback	Same genesis, lower epoch	Skip epoch comparison, warn
Slot rollback within epoch	Same genesis, same epoch, lower slot	Info-level log

The same logic protects other clusters from incorrect epoch summaries during rare restart events.

API cache TTL strategy

Not all data changes at the same rate. SONDA caches intelligently.

Source	TTL	Rationale
Gossip, Validators, Epoch	Always fresh	Core monitoring, never cached
DZ devices, users, multicast	2 min	Real-time infrastructure awareness
BAM validators, nodes	2 min	Fast-changing connections
Trillium	30 min	Updates several times per epoch
Rakurai	1 hour	Validator list changes slowly
BAM IBRL, stake	1 hour	Calculated per epoch
validator-info	4 hours	On-chain, rarely changes
Geolocation	7 to 30 days	Adaptive TTL with jitter, separate SQLite DB
SFDP epoch data	Permanent	Historical data, never changes

Quick Start

Prerequisites

Python 3.10+
Solana CLI in PATH
Optional: doublezero CLI for DZ data
For automation: Cloudflare R2 bucket, Telegram bots, DB-IP and IPInfo accounts

Standalone analyzer (no automation needed)

For a single one-off scan of any cluster:

git clone https://github.com/SolyaUk/sonda.git
cd sonda
pip install requests pyyaml boto3

# Mainnet scan
python3 analyzer/solana_analyzer.py \
    --cluster mainnet-beta \
    --dbip-key YOUR_DBIP_KEY \
    --ipinfo-token YOUR_IPINFO_TOKEN \
    --endpoints analyzer/endpoints.yaml \
    --geo-overrides analyzer/geo_overrides.yaml \
    --export \
    --output /tmp/snapshot.json

# Alpenglow community cluster
python3 analyzer/solana_analyzer.py \
    --cluster alpenglow-community \
    --rpc-url http://YOUR_ALPENGLOW_NODE:8899 \
    --dbip-key YOUR_DBIP_KEY \
    --ipinfo-token YOUR_IPINFO_TOKEN \
    --endpoints analyzer/endpoints.yaml \
    --export \
    --output /tmp/snapshot-alpenglow.json

Historical data collector

For per-validator location history back to epoch 196 (2021):

mkdir -p ~/sonda_data/imports
cd ~/sonda_data/imports

# Initial collection for mainnet (takes around 6 hours, resumable)
python3 ~/sonda/analyzer/solana_history.py \
    --cluster mainnet-beta \
    --dbip-key YOUR_DBIP_KEY \
    --ipinfo-token YOUR_IPINFO_TOKEN

# Resume after interruption (Jito phase re-fetches in 40s, SFDP continues from checkpoint)
python3 ~/sonda/analyzer/solana_history.py --resume

# Incremental update later (add new epochs, preserve all SFDP data)
python3 ~/sonda/analyzer/solana_history.py --update --dbip-key YOUR_KEY

See analyzer/README.md for the full collection-to-R2-upload workflow.

Full production automation

For a continuously-running, multi-cluster, R2-publishing, Telegram-alerting deployment:

# 1. Set up directories
mkdir -p ~/sonda_data/{snapshots,backups,current,history,imports}
mkdir -p ~/sonda/automation/logs

# 2. Configure
cp automation/config.example.yaml automation/config.yaml
nano automation/config.yaml   # Fill in API keys, R2 credentials, Telegram tokens

# 3. Install systemd service (one-time)
sudo ln -s ~/sonda/systemd/sonda.service /etc/systemd/system/sonda.service
sudo cp ~/sonda/systemd/sonda-logrotate /etc/logrotate.d/sonda
sudo chown root:root /etc/logrotate.d/sonda
sudo systemctl daemon-reload
sudo systemctl enable sonda.service

# 4. Enable pre-commit hook (one-time, after cloning)
git config core.hooksPath .githooks

# 5. Start
sudo systemctl start sonda.service
journalctl -u sonda.service -f

Detailed operational instructions: systemd/README.md, automation/README.md.

API keys

Service	Required for	Free tier
DB-IP	Primary geo + ASN lookup	1K lookups/day on free tier (paid Starter plan: 10K/day with Extended API)
IPInfo	Secondary geo verification	50K lookups/month
GeoJS	Tertiary geo (no key needed)	Unlimited
ip-api	Discrepancy resolution (no key needed)	45 req/min
RIPE Stat	ASN name lookup (no key needed)	Unlimited
Cloudflare R2	Public data hosting (automation only)	10 GB storage free
Helius	Recommended mainnet RPC (automation)	1M credits/month on free tier (paid Developer plan: 10M/month)
Telegram BotFather	Notification bots (automation only)	Free

Output Format

Real-time snapshot (`solana_analyzer.py`)

Structured JSON with predictable field schemas per node role.

{
  "timestamp": "2026-05-26T09:55:00.000000+00:00",
  "cluster": "alpenglow-community",
  "epoch": 30,
  "slot": 1641261,
  "epoch_completed_percent": 39.37,
  "genesis_hash": "3QWCajStkp68qAAgCjofJ3BpCyYfPQFxSVZppkYrSpju",
  "features": {
    "total_count": 278,
    "active_count": 278,
    "pending_count": 0,
    "pending": []
  },
  "record_counts": {
    "validator": 82, "rpc": 4, "infrastructure-node": 2
  },
  "records": [
    {
      "identity_pubkey": "...",
      "role": "validator",
      "ip_address": "1.2.3.4",
      "geolocation": {
        "country_code": "US", "city": "Ashburn",
        "asn": "AS24940", "asn_name": "HETZNER-AS",
        "confidence": "high", "discrepancy": false
      },
      "version": "0.3.2",
      "client_type": "Anza Alpenglow (1)",
      "bls_pubkey": "7B34dCYCh9wkUUmBpxrbNNsuU1gxMjRgLBs6PhDCdzMbJeyUu2tgUPkw7zJLLdobuT",
      "stake_percentage": 1.72,
      "is_rakurai": null,
      "dz_connected": false
    }
  ],
  "metrics": {
    "validators": {
      "decentralization": {
        "nakamoto": { "country": 3, "asn": 4, "city": 5, "validator": 9 },
        "hhi": { "country": 1842, "asn": 1201 },
        "gini": 0.847,
        "shannon_entropy": { "country": 0.71, "asn": 0.68 }
      }
    }
  }
}

Three new top-level fields versus older versions: genesis_hash, features, and per-validator bls_pubkey. The first two are populated for every cluster. bls_pubkey is currently populated only for alpenglow-community (SIMD-0387 active there); other clusters will follow once the VAT feature activates.

Historical timeline (`solana_history.py`)

Per-validator datacenter history from epoch 196 (2021) to present.

{
  "meta": {
    "cluster": "mainnet-beta",
    "fetched_at": "2026-03-27T12:26:40Z",
    "sources": ["jito_kobe", "sfdp", "stakewiz"],
    "total_validators": 788
  },
  "epoch_dates": { "500": "2023-02-15", "947": "2026-03-25" },
  "validators": {
    "HwcVgFSg...": {
      "identity": "HwN6eoEe...",
      "vote_account": "HwcVgFSg...",
      "location_changes": [
        {
          "from_epoch": 231, "to_epoch": 499,
          "ip": null,
          "country_code": "CA", "city": "Beauharnois",
          "asn": "AS16276", "asn_name": "OVH",
          "source": "sfdp",
          "dc_stake_percent": 1.79
        },
        {
          "from_epoch": 500, "to_epoch": 804,
          "ip": "15.235.13.5",
          "country_code": "CA", "city": "Montreal",
          "region": "Quebec", "latitude": 45.50, "longitude": -73.57,
          "asn": "AS16276", "asn_name": "OVH", "isp": "Ovh Sas",
          "source": "jito"
        }
      ]
    }
  }
}

Public data on Cloudflare R2

The automation pipeline publishes split files to data.sonda.network on every cycle.

data.sonda.network/
├── current/
│   ├── mainnet-beta/
│   │   ├── network_summary.json     ~100 KB (meta + metrics, no records)
│   │   ├── validators.json          ~1.2 MB (~780 validators)
│   │   ├── rpc.json                 ~130 KB
│   │   └── infrastructure.json      ~160 KB (DZ, BAM, Jito, Harmonic)
│   ├── testnet/
│   ├── devnet/
│   └── alpenglow-community/
├── history/
│   ├── mainnet-beta/
│   │   ├── _index.json
│   │   └── {identity}.json          ~5 KB each, 787 files
│   ├── testnet/                     761 files
│   └── devnet/                      36 files (mostly empty)
└── backups/
    └── timeseries-YYYY-MM-DD.db.gz  Nightly SQLite backups

Production Status

Component	Status	Notes
Real-time analyzer	Live	Running every 60s (mainnet, alpenglow-community), 300s (testnet, devnet)
Historical collector	Complete	Initial pull done April 2026; re-runnable for gap fills
Multi-source geolocation	Live	DB-IP Starter plan + IPInfo + GeoJS + ip-api
Cluster rollback detection	Live	Three-type detection deployed May 2026 with Alpenglow integration
BLS pubkey tracking	Live	Currently `alpenglow-community` only
systemd service	Live	4 workers, auto-restart, daily log rotation
Cloudflare R2 publishing	Live	data.sonda.network, public read
Telegram notifications	Live	Debug bot (private) + SONDA Network Events (public)
Public dashboard Phase 1	Live	sonda.network, all 4 clusters
Per-validator pages	In progress	Phase 2
Per-datacenter pages	In progress	Phase 2

Roadmap

Phase 1 (delivered)

Core analyzer with 4-source geolocation
Infrastructure mapping for BAM, DoubleZero, Rakurai, Jito, Harmonic
API cache with per-source TTL strategy
Geo overrides with full audit trail
Historical collector back to 2021
Resume and update modes for fault-tolerant collection
Multi-cluster automation (mainnet, testnet, devnet, alpenglow-community)
Public dashboard at sonda.network

Phase 2 (in development)

Per-validator pages with field-change timeline and location history
Per-datacenter pages with hosting analytics and incident history
Epoch snapshot auto-push to R2 with full backfill
Cluster-wide time series (Nakamoto, HHI, Gini trends)
Endpoint health uptime API

Phase 3 (planned)

BLS pubkey adoption tracking expanded to all clusters once SIMD-0357 (VAT) activates
RPC fallback chain for the alpenglow-community cluster
Infrastructure health score (synthetic SONDA metric for validator ranking)
Public API access with JSON exports
Attribution program for downstream tools

Phase 4 and beyond

Synthetic Validator Score and Datacenter Score composites
Embeddable widgets, custom alerts, watchlists
Internationalization

Built By

Created and maintained by Solya Validator, an independent Solana validator running since September 2021. Currently hosted in Singapore (AS20473 Vultr/Edgevana, same provider as São Paulo). Relocated in May 2026.

Identity: HwN6eoEe9N3kwHi66hpQDBMFPk6ASQGthWKPX5MZmisp Vote: HwcVgFSgmfeeF7zGFUBLoVA8Hpx8rtwyfCrJ1npBaSVC

Project Twitter: @SondaNetwork Operator Twitter: @SolyaOS

Contributing

SONDA is open source under MIT. Contributions welcome via pull requests. For larger changes, open an issue first to discuss the approach.

When working with the codebase, note the security model: automation/config.yaml contains live API keys and is gitignored. The .githooks/pre-commit hook blocks accidental commits of this file and detects obvious secret patterns. Enable hooks after cloning:

git config core.hooksPath .githooks

License

MIT

Name		Name	Last commit message	Last commit date
Latest commit History 15 Commits
.githooks		.githooks
analyzer		analyzer
automation		automation
systemd		systemd
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md

Folders and files

Latest commit

History

Repository files navigation

SONDA

Contents

Why SONDA?

Who is this for?

What SONDA Analyzes

Network Nodes (5,000+ per scan on mainnet)

Geolocation: 4-source cross-verification

Decentralization Metrics

Infrastructure Intelligence

Repository Layout

Architecture

Real-time pipeline

Cluster rollback detection

API cache TTL strategy

Quick Start

Prerequisites

Standalone analyzer (no automation needed)

Historical data collector

Full production automation

API keys

Output Format

Real-time snapshot (solana_analyzer.py)

Historical timeline (solana_history.py)

Public data on Cloudflare R2

Production Status

Roadmap

Built By

Contributing

License

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Real-time snapshot (`solana_analyzer.py`)

Historical timeline (`solana_history.py`)

Packages