| Server location | Mainland China for users primarily in mainland China; Hong Kong or another nearby region may suit workloads that do not require mainland hosting. | Location-specific provisioning, hardware replacement processes, and clear information about where data is stored and backed up. | Confirm the physical data-center location, service coverage, data-handling terms, and whether the workload requires mainland hosting. |
| Processor and memory | For a general-purpose workload, consider roughly 8–32 CPU cores and 32–256 GB of RAM, adjusting for application and traffic needs. | Hardware setup, operating-system installation, health monitoring, and help diagnosing resource bottlenecks. | Check the exact CPU model, core count, memory type, upgrade options, and whether resources are dedicated to the server. |
| Storage and redundancy | NVMe or SSD storage is a common choice for latency-sensitive applications. RAID 1 or RAID 10 can improve availability, but is not a backup. | Disk monitoring, replacement coordination, RAID configuration when offered, and documented restore procedures. | Confirm usable capacity after RAID, drive replacement terms, backup location, retention, and restore testing. |
| Network capacity | Evaluate the committed bandwidth and traffic allowance, not just the port speed. A 1 Gbps port does not necessarily mean 1 Gbps of guaranteed throughput. | Network monitoring, incident escalation, and clear policies for bandwidth, traffic limits, and DDoS response. | Ask whether bandwidth is shared or committed, how excess traffic is handled, and which routes serve your target users. |
| Availability and SLA | Use a written availability target such as 99.9% per month as a comparison point. At 99.9%, the theoretical monthly downtime allowance is about 43.8 minutes in a 30-day month. | Hardware and network incident response, maintenance notices, escalation paths, and any service credits defined by contract. | Read the SLA exclusions, measurement method, claim process, and whether maintenance windows count toward downtime. |
| 24/7 technical support | Support should be available around the clock for urgent infrastructure incidents, with response targets stated in writing. | Help with server hardware, connectivity, operating-system issues within the agreed scope, and emergency escalation. | Verify support languages, contact channels, severity levels, response targets, and which tasks incur extra charges. |
| Operating-system management | Choose an OS and software stack compatible with the application; define patching frequency and change-control needs before deployment. | Installation, security updates, service monitoring, and configuration support where these are explicitly included. | Confirm supported OS versions, patching responsibility, reboot approval process, and whether application administration is included. |
| Backups and recovery | A planning target may include daily backups with a defined retention period, such as 7–30 days, plus a tested recovery procedure. | Backup scheduling, failure alerts, documented restore steps, and recovery assistance within the agreed service scope. | Check backup frequency, retention, storage location, recovery point and recovery time objectives, and restore fees. |
| Security and DDoS response | Set security controls according to risk, including firewall rules, access restrictions, patching, and a stated DDoS mitigation policy. | Infrastructure-level monitoring and escalation; application security work should be separately defined if required. | Ask what attack sizes and traffic types are covered, whether mitigation is automatic, and how security incidents are reported. |
| Mainland China website requirements | Websites hosted on mainland China infrastructure generally need the applicable ICP filing; commercial internet information services may require additional licensing. | Guidance on hosting-related documentation and coordination, without treating the provider as a substitute for legal or regulatory advice. | Confirm current requirements for the specific website, domain, service type, and hosting location with the relevant authorities or qualified counsel. |
| Best-fit selection | Prioritize mainland location and compliance needs for mainland audiences; prioritize route quality, support scope, recovery, and predictable bandwidth for other use cases. | A managed plan should clearly state included tasks, exclusions, escalation procedures, and recurring and one-time charges. | Request a written specification and SLA, then compare them against workload tests and recovery requirements rather than choosing on hardware figures alone. |