I'm a web developer with more than two years of experience, and I've recently taken on a complete project that includes deployment and infrastructure. The client's existing server is too old for the stack I used: React, Node.js, and PostgreSQL.
The current choices seem to be either buying a new server for roughly $4,500, with licenses bringing the total above $6,000, or using two mini PCs with Ryzen 5 3500U processors, 16 GB of DDR4 memory, and 512 GB SSDs. The mini-PC option, including an automated backup setup, would cost about $1,000.
The application has modest requirements: around 12 concurrent users, a PostgreSQL database of approximately 700 MB, and potentially 1.3 GB of data processed if all users run the heaviest map-related task simultaneously. That processing footprint may be reducible to under 100 MB by generating a static map file.
I doubt a $6,000 server is justified for this workload. Two mini PCs seem capable, with one running the application and database and the other acting as a standby or backup system. My main concerns are recovering data after a failure and avoiding a complete outage if one machine or drive dies. Should I use the mini-PC setup, buy proper server hardware, or host the application on a VPS or cloud platform instead?
5 Answers
The most important missing number is the business cost of downtime. If an outage can wait until you replace a failed machine, inexpensive hardware may be fine. If the application is critical, compare the cost of a managed VPS, redundant hosting, or professional support instead of trying to build a small cluster yourself.
A second mini PC does not replace proper backup and disaster recovery. Plan for at least one local backup and one off-site copy, retain several versions, and perform test restores before launch. Also consider moving the map-heavy processing to the client when practical or generating static map data ahead of time to reduce server load.
Price a small VPS or managed cloud instance before buying hardware. This workload is tiny for modern hosting, and automatic snapshots, off-site backups, and replacing failed virtual hardware are usually easier than maintaining physical machines. You could likely run the application and database on a very small instance, with the frontend served separately if needed.
You probably do not need a $6,000 server for these numbers. A used business-grade server or workstation running Debian or another server-focused Linux distribution would have plenty of capacity. Spend the budget on reliable storage, monitoring, documented recovery procedures, and tested backups rather than excess CPU and memory.
If the application must stay on-site, two mini PCs can work, but they are not automatically high availability. Use identical machines, keep the second as a warm standby, replicate or regularly back up the database, and store encrypted copies off-site. A nightly backup to another drive in the same room will not protect against theft, fire, or a bad update.
Also decide who handles failures at inconvenient times. The hardware may be inexpensive, but the cost of downtime and your long-term responsibility for patching and monitoring can exceed the purchase price.
The main advantage is avoiding a hardware closet, power issues, and being responsible for uptime. Just make sure the backups are separate from the VPS and that you have actually tested restoring one.