4 GB RAM VPS Hosting

Below 4 GB of RAM most modern Linux stacks start fighting the kernel page cache for working set memory. A 2 GB VPS can run a single service but rarely runs an application server, a database and a cache together without swap. At 4 GB you finally have enough headroom for the kernel to keep its filesystem cache warm while your processes hold their working sets in resident memory. That difference shows up as flat tail latency on every request, fewer disk reads for hot data and predictable response times for any service that touches both compute and storage.

4 GB of RAM comfortably hosts a small SaaS MVP with Postgres and Redis on the same VM, a Mailcow installation for a small team, a self hosted Nextcloud for around 20 to 30 active users, a Matrix Synapse server federating with a few hundred other servers, a small Mastodon instance, a private Gitea or Forgejo with a CI runner, a Vaultwarden password manager for a team, a Plausible analytics deployment for several websites and a small Discourse forum. The memory envelope also covers most game server workloads such as Minecraft Java with mods and small Counter Strike 2 servers.

On a 4 GB VPS the easiest memory wins come from tuning Postgres shared_buffers to around 1 GB, MariaDB innodb_buffer_pool_size to around 1 to 1.5 GB and giving the OS another 1 to 1.5 GB for filesystem cache. Avoid running multiple JVM heavy services on a 4 GB box because each JVM tends to reserve hundreds of megabytes for its own bookkeeping. If you must run Java workloads, use Quarkus, GraalVM native image or aggressive container size limits. Anubiz Host exposes detailed memory metrics in the panel so you can see resident, cache and swap pressure in real time.