OpenVPN's documentation for Access Server is one of the few that gives sizing as formulas you can apply, and the formulas, checked on 1 November 2026, are what a team needs to size a VPN gateway for remote staff or for private access to its own environments. This guide applies them and prices three sizes of the OpenVPN Access Server package on MassiveGRID PaaS, built on Virtuozzo Application Platform (formerly Jelastic).
What OpenVPN publishes
The system requirements page gives rules rather than a single box. RAM: start with at least 1 GB and add 1 GB for every 150 connected devices, described as a rough estimate; the IoT example asks for at least 8 GB with split tunnelling and more for full tunnel. CPU: about 12 MHz per 1 Mbit/s of traffic, so a 4-core 3 GHz CPU handles roughly 1,000 Mbit/s, and without AES-NI the calculation quadruples. Disk: 16 GB is typically sufficient for logs, certificates and settings, with log rotation planned. Connections: OpenVPN recommends servicing fewer than 1,000 connections per server, limits internet-routed (NAT) clients to 1,000 per installation, and sets a default maximum of 2,048. Supported systems are Ubuntu 26.04, 24.04 and 22.04 LTS, Debian 13 and 12, and RHEL 10, 9 and 8, on a non-static-compiled kernel.
OpenVPN Access Server on MassiveGRID PaaS: one-click install, about $19.68 a month for the Starter configuration (OpenVPN Access Server node, 1 GiB) at $0.003372 per cloudlet-hour, before volume discounts, in New York, London, Frankfurt or Singapore. 14-day free trial, no credit card. See the sizing table.
Applying the formulas to a company
Take a company with 80 remote staff who connect during the working day and a handful of always-on site-to-site links. Devices: about 100 at peak, so 1 GB plus one more, 2 GB. Traffic: if the tunnel carries only internal applications (split tunnel), peak throughput is tens of Mbit/s and the CPU rule is trivially met; if it carries all internet traffic (full tunnel), budget the office's whole bandwidth, say 200 Mbit/s, which the rule prices at 2.4 GHz of CPU. The connection count sits well under the 1,000 recommendation, so one server is right, and a second is for redundancy rather than capacity.
The device fleet case is different: 1,500 sensors reporting over the VPN is 10 GB of RAM by the formula and more than one server by the connection rule, which is where OpenVPN's own IoT example and its recommendation to stay under 1,000 connections per installation come from.
The three configurations on PaaS and their cost
The package installs OpenVPN Access Server with a public IP, the admin and client web portals and an initial administrator. Cloudlets are 128 MiB RAM plus 400 MHz CPU at $0.003372 per cloudlet-hour, $2.46 per cloudlet-month; Access Server's own licensing (free for two concurrent connections, paid per connection above that) is separate and bought from OpenVPN.
| Configuration | Access Server node | Cloudlets | Per month (730 h) | Fits by the formula |
|---|---|---|---|---|
| Starter | 1 GiB | 8 | $19.68 | A handful of users, split tunnel |
| Standard | 2 GiB | 16 | $39.36 | Up to about 150 devices |
| Production | 4 GiB | 32 | $78.72 | Hundreds of devices, full tunnel |
The figures are ceilings for an always-on server before volume discounts, which a VPN gateway usually is; external traffic is billed per gigabyte on top and is the significant line for a full-tunnel deployment. The PaaS pricing calculator prices other sizes.
CPU, encryption and the cloudlet
The 12 MHz per Mbit/s rule assumes AES-NI; the platform's hosts provide it, so the rule applies as written. A cloudlet carries 400 MHz, so the Standard row's 16 cloudlets carry 6.4 GHz of allocation, enough for the 200 Mbit/s full-tunnel example with room, and vertical scaling means the node uses that CPU only while traffic flows. The traffic itself is the thing to watch: a full-tunnel VPN for 80 people is a few terabytes a month of external traffic in both directions, which is a per-gigabyte line on the invoice that no RAM figure predicts.
Private access to your own environments
The most common reason to run Access Server on the platform is not remote workers but the environments themselves: databases, admin panels and internal tools that should never be on the public internet. An Access Server environment in the same account reaches the others over the internal network, so a database with no public IP is reachable through the tunnel and nowhere else. The private access post sets that up for a remote team, and the TLS add-ons encrypt the database connections inside the tunnel for the reviews that ask.
Frequently Asked Questions
How much RAM does OpenVPN Access Server need?
OpenVPN's system requirements, checked on 1 November 2026, say to start with at least 1 GB and add 1 GB for every 150 connected devices, as a rough estimate. The Starter, Standard and Production configurations on MassiveGRID PaaS allocate 1, 2 and 4 GiB.
How many users can one Access Server handle?
OpenVPN recommends fewer than 1,000 connections per server, limits NAT-routed clients to 1,000 per installation, and sets a default maximum of 2,048; CPU is budgeted at about 12 MHz per Mbit/s of traffic with AES-NI. A company of 80 remote staff is well inside one server.
What does a self-hosted VPN cost per month?
About $19.68, $39.36 or $78.72 a month for the three configurations on MassiveGRID PaaS at $0.003372 per cloudlet-hour before volume discounts, plus external traffic per gigabyte, which is significant for full-tunnel use. OpenVPN's connection licences above the free two are bought from OpenVPN.
A VPN gateway sized by OpenVPN's own formula
The OpenVPN Access Server package on MassiveGRID PaaS installs the server with a public IP and both web portals in one click, from about $19.68 a month, with private access to every other environment in your account. 14-day free trial, no credit card.
OpenVPN Access Server on PaaS