Workload planning

Four workloads for dedicated servers

The 32-core, 64-thread server with ECC memory is dedicated to a single customer.

View monthly pricing for 128GB and 256GB

Start by mapping workloads to resources

All four workload types use the same AMD EPYC 7543P hardware. The differences lie in how memory is reserved, disks are configured, and IP addresses and bandwidth are allocated—not in having a separate server model for each workload.

Virtualization host

Run VMs and containers on Proxmox VE

Install Proxmox VE 9 and use ZFS mirroring for VM disks. Additional IPv4 addresses can be purchased and assigned to VMs that need their own address.

Reserve memory first for the host and the ZFS cache, then allocate the rest to KVM VMs or containers. Set CPU quotas based on concurrent workloads rather than simply adding up the virtual core counts assigned to each VM.

The two included NVMe drives can be configured as a mirror during installation. Mirroring helps protect against a single-drive failure, but does not replace off-site backups. VM templates, configuration files and business data should each have a recovery plan.

Plan for: host reserves and VM isolation

Databases & middleware

Run the primary database and services on one server

Run a PostgreSQL or MySQL primary on NVMe, with Redis and message queues on the same server. The 256GB configuration gives more room to plan for buffer pools.

Memory planning should account for the database cache, connection processes, the operating system cache and middleware. Do not allocate all memory to the buffer pool or estimate capacity based only on database file size.

When the primary database, logs and queues share physical disks, check whether their write peaks overlap. Direct-attached NVMe does not eliminate I/O contention between workloads. Critical services need resource limits, monitoring and tested backup recovery.

Plan for: cache and write peaks

Websites & APIs

Host website entry points and application services together

Use a dedicated IPv4 address for websites and APIs. Choose 250 Mbps, 1 Gbps or 2 Gbps based on outbound traffic needs; all three options are unmetered.

The server is located in Tokyo, Japan. Visitor experience also depends on network conditions, international routes, page asset sizes and application response times. A server's location does not guarantee the same latency for every visitor.

Resources for multiple websites can be separated by process, container or VM. Whether additional dedicated IP addresses are needed depends on workload isolation and network design; the number of websites alone does not determine how many addresses to buy.

Plan for: outbound peaks and visitor locations

Windows workloads

Business software and Hyper-V hosts

Windows Server Datacenter can run business software and Hyper-V. Teams connect over RDP to manage the server, and the operating system licence is billed monthly with the server.

Windows Server 2019, 2022 and 2025 are available in English, Traditional Chinese or Simplified Chinese. Check application and database software compatibility and licensing terms separately before deployment.

RDP is a management access method; it does not include business software licences or additional remote-use rights. For deployments with multiple VMs, check the licensing terms for each guest operating system and reserve resources for host management and backups.

Plan for: compatibility and licensing

Allocate memory for working sets and concurrency

128GB is a starting point for a single database or a few VMs. 256GB provides more room for planning a dozen or more VMs or a database with a larger buffer pool. These figures are planning guidance, not a guarantee of how many instances can run.

VM density depends on each guest's resident memory, concurrent workloads and peak demand. For databases, estimate requirements from the active dataset, connection count and query patterns—not just disk usage.

Both configurations use DDR4-3200 ECC memory and have the same CPU, included drives and included bandwidth. Memory cannot be upgraded on the original server after provisioning; changing configurations requires ordering another server. Plan for both steady-state and peak resource needs before choosing.

Choose an environment based on resource boundaries

Shared VPS hosting, a dedicated server and a self-managed rack involve different levels of management and investment. This table compares typical characteristics only: monthly server rental does not include application management, and a dedicated server still requires system administration.

Hosting options compared
Category	Shared VPS	Dedicated server	Self-managed rack
Resource isolation	Uses allocated virtual resources; the underlying physical hardware typically hosts multiple customers.	The entire server is dedicated to one customer. CPU, memory and drives are not shared with other customers.	Resource ownership depends on the purchased equipment and rack deployment.
Compute consistency	Depends on virtualization scheduling, resource quotas and sharing on the underlying hardware.	The 32 cores are not shared with other customers; workloads on the same server still need resource planning.	Depends on the hardware and workloads on the server; equipment operation is self-managed.
Provisioning	Provisioned according to the provider's virtual instance configuration and process.	Ordered online and provisioned through the control panel. The customer manages the operating system and application deployment.	Equipment is purchased, racked and cabled independently, with operating system and network deployment managed in-house.
Initial cost	Typically starts with a smaller resource allocation, with pricing based on the selected plan.	Monthly rental, no setup fee and no contract.	Includes equipment, rack space, networking and on-site management costs.

Dedicated resources refer to physical hardware, not guaranteed application performance. Database lock contention, application configuration, backup writes and VM overcommit can still create bottlenecks on the same server.

Network planning: bandwidth and IP addresses

250 MbpsIncluded · unmetered

1 GbpsOptional · unmetered

2 GbpsOptional · unmetered

IPv4 allocation order

First address: Host management and SSH / RDP access
Additional addresses: Assigned to workloads or VMs as planned
Total addresses: 1 included, up to 256

This is a resource-planning example, not a display of connection credentials. Assigned IP addresses are available in the control panel after payment and provisioning.

Choose bandwidth for peak demand, IPs for isolation

Bandwidth determines the available transfer rate; traffic describes the total amount of data transferred. All three options are unmetered, but their speeds differ. Include overlapping file distribution, website traffic and backup transfers in outbound capacity planning.

Start by reviewing application response data, file downloads and other outbound workloads during peak periods, then choose 250 Mbps, 1 Gbps or 2 Gbps. Bandwidth is a single selectable tier; the speeds of the bandwidth options should not be added together.

One dedicated IPv4 address is included. Up to 255 more can be purchased individually. Multiple websites and VMs can be assigned addresses based on isolation needs. When multiple websites use one address, application entry points and certificate configuration are also required.

Multiple IPv4 addresses do not automatically configure guest networking. The first address in the host's allocation list is reserved for SSH or RDP. Configure the remaining addresses in the operating system or VM using the provided network parameters, and review firewall rules and access permissions.

Plan data protection before deployment

The included 2×1TB NVMe drives are independently connected, with no hardware RAID. Their combined raw capacity is 2TB; this does not mean they are already configured as RAID 1. The file system, operating system usage and redundancy method all affect usable capacity.

For database and VM workloads, configure software RAID or a ZFS mirror first, then set up off-site backups. Mirroring provides redundancy at the disk layer, but cannot protect against accidental deletion, application-level data corruption or damage to data on the same server.

Three checks before going live

Disk layout: Decide whether to use the drives independently or as a mirror, and define where application volumes, logs and backup files will reside.
Recovery sources: Keep backup copies off the server. Database backups should match the chosen recovery method.
Recovery testing: Verify that backups can be read, configurations rebuilt and services started—not just that backup jobs completed.

Hardware and networking are managed by the provider; the customer is responsible for operating system configuration, business data and backups.

Configure a server for the workload

Choose a memory configuration first, then plan for storage, bandwidth, IPv4 addresses and the operating system. Monthly pricing for both configurations and optional extras is listed on the pricing page. The configuration page shows the full quote and amount due for the current billing period.

Prices are settled in USD. Available payment methods are listed on the configuration page. Current availability is shown in real time in the control panel.

Configure server Virtualization host details Windows Server licensing