Block storage vs local disk on a VPS: when to add a volume and what it costs

Every VPS plan comes with a disk, and sooner or later a database, an upload folder or a pile of backups outgrows it. At that point most providers offer two ways forward: attach a block storage volume, a network disk billed per gigabyte that you mount next to the system disk, or resize to a larger plan with more local storage built in. The two look interchangeable in a control panel, but they behave differently under load, fail differently, and are priced on completely different logic. Knowing which one you need before you hit 100% usage is much cheaper than finding out during an outage, and it is worth <a href="/#calc">sizing the plan with disk in mind</a> from the start. This guide covers how local disk and block volumes differ, what they cost, and when each one is the right call.

Local disk and block volumes are different things

The disk that ships with a VPS plan usually sits on fast NVMe in or near the physical host that runs your server. It is the system disk: the operating system, your applications and, on most small setups, the data all live on it. It is fast because the path from CPU to storage is short, and its size is tied to the plan, so on a typical lineup the disk grows together with RAM and cores. If the storage type itself is the question, our guide on NVMe vs SSD for VPS storage explains what the drive generation changes in practice.

A block storage volume is a separate virtual disk delivered over the provider's storage network. You create it at the size you need, attach it to a server, format it and mount it like any other drive. Its strengths are flexibility and durability: you can grow it without touching the server, detach it and attach it to another machine, and the data is usually replicated. Hetzner, for example, documents that each block of a Cloud Volume is stored on three separate physical servers, that a volume can be sized from 10 GB to 10 TB in 1 GB steps, and that a server can carry up to 16 volumes, attached to one server at a time. The trade-off is that every read and write crosses the network, so performance is capped: Hetzner lists up to 5,000 IOPS and 200 MB/s sustained per volume, with bursts to 7,500 IOPS and 300 MB/s.

What a volume costs compared with a bigger plan

Block storage is priced per gigabyte per month, and the rates are public, which makes the comparison easy to run. DigitalOcean charges $0.10 per GiB per month for Volumes, billed hourly for as long as the volume exists, whether or not it is attached to a server. UpCloud lists Standard block storage at $0.097 per GB today and has published a new rate of $0.131 per GB from November 16, 2026, with its high-performance MaxIOPS tier at $0.250 per GB. So a 200 GB volume runs about $20 a month at DigitalOcean, about $19.40 on UpCloud Standard today and about $26.20 after the November change.

Now compare that with simply moving up a plan. On Hostman's published cloud server list, 2 vCPU with 4 GB RAM and 80 GB NVMe costs $10 a month, and 4 vCPU with 8 GB RAM and 160 GB NVMe costs $22. The step up costs $12 and adds 80 GB of local NVMe plus twice the RAM and cores. If you also need the compute, the bigger plan is the cheaper way to get the disk; if you only need space, and especially a lot of it, a volume lets you buy exactly that without paying for cores you will not use. For workloads where both disk speed and size matter, our database VPS picks lean toward plans with generous local NVMe.

Which data belongs where

Put latency-sensitive, write-heavy data on local NVMe: the database files of a busy application, a search index, a build cache. These workloads do many small random reads and writes, and a network volume with an IOPS ceiling becomes the bottleneck long before it runs out of space. Put bulky, growing and less frequently touched data on a volume: user uploads, media libraries, logs you keep for months, local copies of backups. Volumes also suit anything you may want to move to another server later, because detaching and reattaching is faster than copying hundreds of gigabytes across the network.

One trap catches many people: a volume is not a backup. Replication protects against a failed drive, not against a deleted table or ransomware, and some providers exclude volumes from server snapshots entirely. Hetzner's documentation states that it does not provide backups or snapshots for volumes and that server backups and snapshots do not include attached volumes. Check your provider's rules before you rely on them, and follow the approach in our guide on VPS backups and snapshots: at least one copy of important data outside the server and outside the volume it lives on.