Airline information systems are among the most integration-heavy estates in any industry: a single journey touches reservation, pricing, inventory, departure control, baggage, loyalty, government APIs, and partner airlines—often on different platforms and continents. IBM MQ has connected these components for decades because flight events are time-sensitive yet cannot be dropped when a downstream system lags during a holiday booking surge or a weather cancellation. This tutorial describes reference flows for PNR create and change, inventory and availability request/reply, operational pub/sub for delays and gate changes, codeshare partner channels, and irregular operations (IRROPS) messaging patterns beginners can map to queues, topics, and channels in a diagram.
The reservation host (often a mainframe or packaged reservation system) remains the system of record for bookings. Web and mobile channels send booking requests to an integration tier that MQPUTs to RES.BOOKING.REQUEST. A host adapter GETs messages, calls host APIs or transactions, and PUTs confirmation or error to RES.BOOKING.RESPONSE with CorrelId. Peak sales (fare sales, holidays) produce depth on the request queue; auto-scaling consumer instances compete for GET under syncpoint. MAXDEPTH and consumer count must be load-tested with realistic message sizes—PNR XML can be large, affecting channel BATCHSZ tuning.
123456DEFINE TOPIC('FLIGHT.OPS') TOPICSTR('FLIGHT/OPS/>') DEFINE SUB('SUB.GATE.DISPLAYS') TOPICSTR('FLIGHT/OPS/GATE') + DEST('OPS.GATE.Q') DESTQMGR('QM_HUB') DEFINE QLOCAL('OPS.GATE.Q') MAXDEPTH(100000) DEFPSIST(YES) * Topic string hierarchy — subscribers use wildcards for branch events * DEST queue receives publications matching subscription
Availability checks are classic request/reply: shopping engines cannot hold inventory locks on every search, but booking must confirm seats atomically at purchase time. Search may hit a cache; book sends a correlated message to INVENTORY.HOLD with timeout. If inventory service is slow, shoppers see timeout while messages remain—operations watch hold queue depth. Overbooking protection and cabin upgrades publish events to topics consumed by revenue management systems.
| Event | Pattern | Latency need |
|---|---|---|
| Book segment | Request/reply | Seconds |
| Cancel segment | One-way + ack topic | Seconds |
| Gate change | Pub/sub | Sub-minute |
| Baggage scan | One-way high volume | Seconds |
| Crew roster update | Pub/sub durable | Minutes acceptable |
Departure control systems consume passenger check-in and boarding events. Messages may flow from airport kiosks through a local spoke queue manager to the hub. Weight and balance systems receive aircraft load updates. These feeds mix persistent critical messages with higher-volume scans. Separate queues prevent boarding scans from blocking ticketing: OPS.BOARDING.SCAN versus RES.TICKET.ISSUE. Priority on MQMD can elevate crew deadline messages when supported end-to-end.
Partner traffic crosses organizational boundaries with dedicated channels, mutual TLS, and strict CHLAUTH. Messages often use industry formats (EDIFACT, XML) agreed between carriers. A partner hub puts to PARTNER.INBOUND; your adapter validates and maps into internal PNR messages. Latency on WAN channels affects customer experience—monitor channel status and round-trip times. Never share one MCAUSER across unrelated partners.
Cancellations and diversions generate storms of pub/sub events: FLIGHT/OPS/CANCELLED/BA123. Mobile apps, SMS gateways, hotel voucher systems, and crew schedulers subscribe. Work queues IRROPS.REBOOK.WORK hold per-passenger tasks agents complete. Depth here directly maps to call center wait times. Architects plan for 10x normal pub rate during hub closures; topic performance and subscriber queue depth are drill scenarios.
Pub/sub is the airport loudspeaker: one announcement reaches every gate display and crew radio that subscribed. Point-to-point queues are phone calls between two offices—booking talks only to the host adapter. You need both in an airline.
Many carriers still run host inventory on z/OS with MQ channels to Linux integration hubs. The mainframe integration tutorial covers CICS; airlines often mirror the same hub pattern as banks—z/OS as spoke or hub depending on era of migration. Cloud-native digital retail for ancillaries (bags, seats) may use separate queue managers with cluster channels into the enterprise hub.
When you book a plane seat, your computer sends a note to the airline office. IBM MQ is the tube that carries notes between offices—ticket office, airplane door, big screen at the gate. If one office is busy, notes wait in a box until someone reads them. When the gate changes, one announcement goes to every screen that is listening.
Design topic strings for gate change, delay, and cancel for flight BA123.
Split high-volume baggage scans and booking requests into separate queues; justify why.
Write IRROPS runbook: topic publish rate 5x normal for 30 minutes—what do you check?
1. Gate change notifications suit:
2. PNR data lives in:
3. IRROPS volume spikes need:
4. Seat availability inquiry often uses: