DevOps for Logistics & Mobility — always-on platforms

What regulates the infrastructure

What actually goes wrong here

• Terminal operating system or warehouse management system outage stops physical movement even though the cranes, trucks and yard are fine: without the system the yard has no authoritative container position and no gate authority, so recovery means manual tallying and re-inventorying thousands of boxes, which is why restarts take days rather than hours.
• EDI and customs message broker failures cause missed statutory filing windows (ICS2, NCTS, e-AWB). The damage is deadline-shaped rather than availability-shaped — an hour of downtime at the wrong moment means refused loads and detained cargo, while the same hour overnight costs nothing.
• Ransomware propagating across an acquisitive, merged IT estate. Freight forwarders and carriers typically run many acquired legacy systems on flat networks; NotPetya reached 17 Maersk/APM terminals in 2017, with damage the company put at USD 250-300 million, and the sector's consolidation pattern has not fundamentally changed.
• Peak-day capacity failure in sortation, label generation and track-and-trace. These systems are sized against average volume and then hit with several times that on a single named shopping day, and a failed label print run halts a physical sort line immediately.
• Loss of telematics and cold-chain telemetry. For temperature-controlled and pharmaceutical freight a monitoring gap is not just missing data — an unbroken temperature record is often the condition of the cargo insurance claim and of GDP compliance, so the consignment can be rejected on the record alone.
• Concentration through a single 3PL or port community system: because shippers and carriers share one platform per port or corridor, an outage at that one operator freezes many independent companies at once, and none of them can fail over because there is no second instance of that port's community system.

How demand behaves

Demand is event-driven rather than steady. Parcel and warehouse volumes concentrate around Black Friday/Cyber Monday, Singles' Day, the pre-Chinese-New-Year shipping rush followed by the factory shutdown lull, and the December peak. Within a day, load is bimodal: a morning scan/dispatch burst and a late-afternoon spike as carriers approach their booking and manifest cut-offs, so a system sized on daily averages will still fail at the cut-off.

Data you will be holding

Consignment records hold shipper and consignee personal data (GDPR where EU persons are involved), plus commercially sensitive cargo descriptions, declared values and routing — information that is itself a theft-targeting asset for high-value freight. Customs filings carry EORI/AEO identifiers and tariff classifications whose accuracy has direct legal consequence.

Architecture this pushes you toward

Estates are unusually heterogeneous — a forwarder's stack typically spans a transport management system, warehouse management, customs filing middleware, EDI/AS2 or API integrations to hundreds of trading partners, and telematics ingest, often across acquired subsidiaries running different vendors. Edge presence matters physically: scanners, gate kiosks and handheld terminals at depots need to keep operating when the WAN link drops, which pushes designs toward local buffering with later reconciliation. Integration surface, rather than raw compute, is usually the dominant complexity and the dominant attack surface.

Availability expectation

Not usually expressed as a single industry figure. 3PL and carrier contracts typically define availability per system in commercial SLAs, while port community systems, terminal operating systems and national customs interfaces run continuously because vessel and truck arrivals are not office-hours events. Treat any single stated percentage as contract-specific rather than an industry norm.