Building High-Availability Enterprise Clouds: Achieving Cross-Datacenter Active-Active Resilience and VMware Replacement

Under a group-wide operational model, multiple listed subsidiaries of Xinyi Glass share core IT resources to achieve efficient business synergy. This imposes higher requirements on the reliability and security of the IT infrastructure: a maximum fault tolerance time of 15 minutes for core business systems, and the creation of a reliable, high-availability disaster recovery solution to mitigate datacenter-level failures. To meet these requirements, Xinyi Glass constructed a "hybrid cloud architecture." Based on a private deployment of a leading public cloud solution, the company established a dual-datacenter architecture across its Mobile Managed IDC and R&D Center IDC. This architecture utilizes vSphere combined with Dorado all-flash storage to ensure redundancy at the data layer.

However, the hybrid cloud architecture failed to meet actual requirements for core databases in high-concurrency and high-pressure scenarios. Implementing an active-active transformation posed both technical dilemmas and significant financial burdens: the private deployment of public cloud solutions involved high maintenance costs and required proportional storage expansion for cross-datacenter active-active scenarios; meanwhile, active-active solutions based on vSphere and centralized storage, while meeting performance needs, required massive hardware investment and faced rising costs due to VMware's shift to subscription-based licensing. After thorough research and rigorous stress testing in actual environments, Xinyi Glass ultimately chose to build a datacenter active-active cluster using the SmartX ECP, establishing a high-availability IT infrastructure with lower investment costs.

Xinyi Glass deployed a 3-node SmartX ECP infrastructure cluster based on hyper-converged architecture at both its R&D Center and Mobile IDC. By utilizing a public cloud virtual machine as the witness node, a 3+3+1 node active-active cluster architecture was formed to support the databases of HR, Finance, and OA application systems, achieving active-active disaster recovery for databases. When a single physical node in the preferred availability zone (R&D Center) fails, virtual machines (VMs) trigger HA normally, migrating to other nodes within the same zone to avoid network configuration changes caused by cross-site migration. In the event of a datacenter-level failure, business operations can migrate across datacenters, with testing showing that VMs can restart within 3 minutes.

Xinyi Glass also utilized the SmartX ECP to fully replace the previous "vSphere + all-flash centralized storage" architecture that hosted non-active-active business databases. This move supports application system databases such as Treasury Management and TSM while reducing VMware virtualization costs. On the hardware side, the solution employs third-party servers to form a 3-node cluster. Equipped with an all-SATA SSD configuration, the setup effectively meets all performance requirements.