6 simultaneous FAA data feeds. 7,000+ aircraft. 30,000+ data points per second. 5 native C++ graph procedures. All running on less than 1% of one CPU core. This is what a purpose-built graph database can do.
Every number on this page is measured live from the running demo. Not benchmarks. Not theoretical maximums. Actual production throughput.
xrayGraphDB processes 30,000+ individual data fields per second — every ADS-B position field, every SWIM XML element, every flight plan update — ingests them into the graph via binary wire protocol, runs 5 native C++ procedures, computes predictions and deviations for 7,000 aircraft, and streams results to your browser at 60fps. The entire pipeline uses less than 1% of one core.
This demo consumes more real-time FAA data than most commercial aviation applications. Every feed is a live connection processing data right now.
Primary aircraft surveillance. Each of 120 edge workers fetches a different US region every 60 seconds from a different IP — distributed rate limiting. ~7,000 aircraft with position, altitude, speed, heading, squawk, and 25+ fields per update.
Filed flight plans for every IFR aircraft in the NAS. Origin airport, destination airport, filed route, aircraft type, ETA, flight status. ~2,000+ flights tracked. This is how we know where every commercial flight is going — not guessing from the nearest airport.
Authoritative FAA surveillance data. FIXM NAS 3.0 format with position updates, flight plan amendments, arrival/departure events, sector handoffs, controlling ARTCC. ~4,500+ flights with FAA-tracked position data. Bulk sync messages up to 133KB with hundreds of records.
Arrival sequencing and metering. ETA and STA (Scheduled Time of Arrival) for every inbound flight at 300+ airports. Meter fix assignments, TRACON sector routing, ARTCC coordination. This is how the FAA sequences arrivals — and we see every update.
TRACON radar tracks, surface movement position reports, tower departure events, pre-departure clearances. Precise terminal area tracking within ~50nm of every major airport. 28,000+ messages per minute.
Real-time Notices to Air Missions via AIXM 5.1. Temporary Flight Restrictions (TFRs) with circle/polygon geometry render as overlays on the map. Aircraft inside active TFR geometry are flagged as NOTAM violators in real time.
All 6 feeds connect via Solace PubSub+ JMS queues to the FAA's SWIM Cloud Distribution Service (SCDS). Messages arrive in real time — not polled, not cached. The XML is parsed, enriched, and written to the graph in a single tick cycle.
This isn't a flight tracker with a database behind it. Every calculation — prediction, deviation scoring, emergency detection, NOTAM violation checking, arrival sequencing — happens inside xrayGraphDB as native C++ procedures scanning the graph directly.
ADS-B + 5 FAA SWIM
30K data pts/s
30 props/node
xrayProtocol 46K QPS
5 native procedures
WebSocket 60fps
One call returns 28 fields: position, TFMS flight plan, TBFM metering, prediction state, deviation scoring, emergency status, landing detection. Scans :Aircraft nodes directly — no Cypher parsing overhead.
Finds every aircraft squawking 7500 (hijack), 7600 (lost comms), 7700 (emergency), or 7777 (military intercept). Scans all Aircraft nodes in microseconds. Flashing ring + anomaly panel alert.
CALL xray.arrival_board('KLAX') returns every inbound flight sorted by altitude — closest to landing first. Origin, ETA, STA, deviation, aircraft type. All from the graph, not external APIs.
One call, one scan, 13 aggregate metrics: total aircraft, commercial/private/military split, altitude bands, landing count, emergency count, average deviation. The entire NAS summarized in microseconds.
Point-in-circle haversine test: for every TFR stored as a :TFR node, check every :Aircraft node against the geometry. Aircraft inside active TFRs at the correct altitude get flagged. Native C++ graph scan — no PostGIS, no external geometry engine.
PROJECT_3D, HAVERSINE, DEVIATION_SCORE, BEARING, BANK_ANGLE, ENVELOPE_SCORE, VELOCITY_3D, ANGULAR_DIFF, and 280+ more. Native C++ functions executing in microseconds. The math happens inside the database, not in application code.
Every other graph database speaks Bolt — a row-based text protocol designed for interactive queries. xrayGraphDB speaks its own binary columnar wire protocol purpose-built for machine-speed graph operations.
No SQL parsing, no query planning, no AST construction. BULK_UPSERT_NODES sends raw binary property arrays directly into the storage engine. Port 7689. Each aircraft = 30 properties written in a single binary frame.
A hash of each aircraft's state is compared to the last write. Only changed aircraft get written — typically 60-80% are unchanged per tick. This means 7,000 aircraft tracked but only 2,000-3,000 graph writes per cycle.
Known aircraft hex IDs are cached as direct vertex pointers. After the first lookup, subsequent updates go directly to the storage slot — zero skip-list scan, zero index lookup. O(1) per upsert.
Every aircraft is scored by how well reality matches prediction. The color tells you instantly.
When data goes stale, the system projects forward and shows how much of what you see is prediction. Trail blips mark actual ADS-B hits; gaps between them = estimation.
Highly predictable — on airways/STARs, constrained by published procedures. Even descending through FL180 is still predictable. Takes 5 full minutes to reach 99%.
Below 10K: approach procedures, ATC vectors, 250kt speed restriction. More variables, less certainty. 2.5 minutes to 99%.
Base turns, go-arounds, pattern work, VFR traffic. Estimate climbs fast. Scaled by speed — a Cessna at 80kt is harder to predict than a 737 at 250kt.
The system monitors every aircraft's transponder squawk code in real time.
Flashing red ring. Highest priority alert. Auto-zoom on detection.
Flashing orange ring. Aircraft has lost radio contact with ATC.
Flashing red ring. Medical, mechanical, fuel — any declared emergency.
SWA2483 flies Dallas to Denver every day. After a few flights, the system knows what "normal" looks like. The first flight IS the path. Each subsequent flight widens the envelope with observed variance.
Initial tolerance: 2nm lateral, 1,000ft vertical (matching US radar separation). Widens only with evidence from repeated flights.
Tick deviation: 10-second prediction accuracy. Envelope deviation: How does this flight compare to all previous flights with the same number?
Late departures = speed increase (dampened). Weather reroutes = lateral offset with same heading (dampened). Heading reversal on a commercial flight = always flagged.
Search by callsign, registration, or hex. Filter by carrier type, altitude band, deviation color, departure airport, arrival airport. Combine filters freely.
Full detail panel: flight plan (origin/dest/route/ETA from TFMS), instruments (bank/speed/alt), position confidence, deviation scoring, SWIM enrichment data.
Top ribbon shows live: aircraft count, data points/sec, CPU%, RAM, QPS, tick latency. Bottom panel: per-component timing, write path, SWIM message totals.
Per FAA SWIM Terms of Service, aircraft on the Limiting Aircraft Data Displayed (LADD) list are completely excluded from display, storage, and all data processing. The LADD list is sourced from the FAA's NAS Aeronautical Data Exchange (ADX) portal and updated within 5 business days of each monthly publication.
This application consumes data from the FAA System Wide Information Management (SWIM) program via the SWIM Cloud Distribution Service (SCDS). Per FAA Terms of Service: SWIM data obtained via SCDS is not intended for NAS-impacting usage. This data shall not be used as the sole source for any aviation-safety related, law enforcement, or judicial activity that relies on the availability, validity, timeliness, or accuracy of data. Per HR 302 FAA Reauthorization Act of 2018, Section 566, redistributed information may not be characterized as FAA data.
This is a technology demonstration of xrayGraphDB's real-time graph processing capabilities. It is NOT an air traffic control system, NOT a flight safety tool, and NOT a substitute for official FAA systems. Do not make any aviation, safety, operational, or navigation decisions based on this demonstration. Aircraft positions, predictions, and deviation scores are for demonstration purposes only.
ADS-B positions: Airplanes.live crowdsourced ADS-B network via 120 Cloudflare edge workers.
TFMS (R14 Flight Data): FAA SWIM — flight plans, origin/destination, ETA, flight status.
SFDPS (Flight FIXM): FAA SWIM — authoritative surveillance, flight plan amendments, handoffs.
TBFM (Metering Publication): FAA SWIM — arrival sequencing, ETA/STA, meter fix assignments.
STDDS (Terminal Data): FAA SWIM — TRACON tracks, surface movement, departure events.
AIM/FNS (NOTAMs): FAA SWIM — TFRs and airspace restrictions with geometry (AIXM 5.1).
LADD compliance: FAA NAS ADX portal — monthly block list for privacy-protected aircraft.
Aircraft on the FAA LADD (Limiting Aircraft Data Displayed) list are excluded from all display, storage, and processing per FAA SWIM Terms of Service. General aviation and on-demand aircraft registrations on the LADD list are never shown, even in historical data. For LADD program inquiries: LADD@faa.gov or (202) 267-0346.
7,000+ aircraft. 30,000+ data points/sec. 6 FAA feeds. 5 native procedures. Less than 1% CPU. All inside xrayGraphDB.
Launch Live Demo