Match by workload
A server is chosen not by gigabytes but by what will run on it. Every page has a configuration with an explanation of where the numbers come from and a list of what to check with the provider before paying
VPS for a website or blog
Hosting for a website, landing page, portfolio or WordPress blog
A website rarely runs short of CPU: the bottleneck is almost always memory and disk operations
VPS for an online store
A store on WooCommerce, OpenCart, Bitrix or a custom CMS
A store differs from a website in that not everything can be cached: the cart, the account and the checkout are computed on the fly
VPS for 1C and Bitrix
1C:Enterprise, Bitrix24 and corporate systems
The requirements of 1C depend not on the number of employees but on the database’s operating mode and the number of concurrent sessions
VPS for a Telegram bot
Bots, webhooks and small always-on services
A bot is almost always idle: it waits for an event, replies in tens of milliseconds and waits again
VPS for a database
PostgreSQL, MySQL, ClickHouse or Redis under load
For a database, memory is not headroom but a working tool: the more data fits in the cache, the fewer trips to disk
VPS for AI and ML workloads
Model inference, n8n with AI, vector databases, your own LLM front-end
AI tasks on a virtual server usually mean not training models but inference, orchestration and vector search
VPS for a game server
Minecraft, CS2, Rust and private game servers
For a game server, single-core frequency and network latency matter most, not the number of cores
VPS for development and testing
Staging, CI/CD runners, sandboxes and dev environments
For development, flexibility matters more than power: a server is spun up for a task, lives a week and is deleted
VPS for n8n and automation
Self-hosted n8n, Coolify and process automation
Self-hosted automation is bound not by computation but by memory: each active process holds an execution context, and the task queue accumulates in the database