A cloud campus in Sincan, west of Ankara, curtailed services for seven hours and 12 minutes on 18 September after floodwater disrupted the supply used to cool its servers. Poyraz Compute, the fictional operator of the facility, said 2.4 million customer sessions were affected; it reported no loss of stored data.
The outage followed 94 millimetres (3.7 inches) of rain in 75 minutes, after three days when temperatures reached 38°C (100°F). The company’s incident bulletin said silt in the municipal supply forced the campus to draw on its limited reserve, then reduce computing loads to keep equipment within safe operating temperatures.
A chokepoint in the cloud
That sequence matters beyond one company. Cloud services support payment processing, online retail, business software and government portals, yet customers often cannot see which physical site handles their requests. When a facility cuts capacity, users may see failed transactions or blank screens even if their own phone and internet connection are working.
“We have designed the cloud as if distance were a backup. It is not,” said Hortensia Dinu, a network-resilience researcher who reviewed the operator’s public incident bulletin. She said a second data centre only provides meaningful protection if it has separate power, cooling and network routes, and can take over quickly.
We have designed the cloud as if distance were a backup. It is not.
Hortensia Dinu, network-resilience researcher
Cooling reserves met the flood
Poyraz said its cooling plant normally uses treated water to carry heat away from server rooms. The flood did not enter the building, but heavy sediment led the municipal supplier to suspend deliveries while it checked its equipment. The campus switched to stored water and then limited non-essential workloads, according to the company.
Léon Guyot, a water-systems engineer who advises data-centre operators, called that a foreseeable design risk. “A site can have two power feeds and still be one broken water connection away from a shutdown,” he said. He warned that hotter operating conditions leave less time to respond when cooling supplies falter.
The episode also puts pressure on a business model built around large, steady electricity and cooling demands. Poyraz said it plans to spend ₺1.8 billion (about €42 million) on a second Ankara-region facility and separate water storage. The work is due to finish in 14 months, but the company has not said whether the new site will be far enough away to avoid the same local disruption.
Francisco S. Castro, a technology-risk analyst, said Turkey’s push to host more digital services at home could deepen the consequences of outages if demand grows faster than backup capacity. “Local data storage is not the same thing as resilient data storage,” he said, urging operators to publish clearer figures on how much work their backup sites can absorb.
Poyraz said it has begun testing a transfer of customer workloads to a second region and will publish the results before the end of the year. Until then, the Sincan outage leaves a blunt lesson for businesses and public agencies: a service that appears to live in the cloud may still depend on one vulnerable pipe, in one place, on one very wet afternoon.
