How Suning Scaled Cloud Resources for the 818 Mega‑Sale: 10% Server Utilization Boost

Improving Resource Scheduling Efficiency – 10% Physical Machine Utilization Increase

In the private‑cloud domain, Suning Cloud introduced a micro‑scheduling product to raise overall physical‑machine usage by 10% during the 818 promotion while maintaining stability and reducing cost.

By refining monitoring granularity to seconds, applying distributed monitoring and edge‑computing principles, they built resource profiles, trained predictive models, and established health scores for resource contention and usage, enabling rapid PCPU‑level micro‑scheduling and QoS‑driven adjustments.

Ensuring Stable Resource Allocation for Hundreds of Thousands of VMs

The 818 event generated traffic dozens of times higher than normal. Suning Cloud used dynamic resource scheduling to precisely allocate resources to hundreds of thousands of virtual machines, avoiding over‑commitment conflicts and keeping physical‑machine load within safe limits.

Capacity planning captured application load profiles across shopping, payment, logistics, and other services, and combined them with physical‑machine capabilities to match resources accurately. Real‑time metrics (CPU, memory, network, disk I/O) guided secondary scheduling and adjustments, with simulation‑based placement when resources were scarce.

Future work will deepen load‑profile usage, applying AI algorithms to predict resource demand across promotion phases, achieving finer‑grained, time‑shared scheduling while integrating kernel isolation to lower competition and improve utilization.

Building a Comprehensive Security System for the 818 Promotion

With increased hacker activity during the promotion, Suning’s security team deployed extensive vulnerability scanning, risk management, and expert services, scanning thousands of systems and domains before the event.

Using SDK and KMS, data is protected throughout its lifecycle. The database audit system secures data assets. An AI‑driven web‑attack detection engine, leveraging big‑data and NLP, achieves 98% accuracy and blocks over 1 billion attacks, generating attacker profiles and pre‑emptive blocks.

Overall, Suning Cloud delivered high stability, scalability, and availability during the 818 peak, and will continue post‑event analysis to further optimize performance.