I built a website with a search feature and want to add autocomplete suggestions like major search engines do. Instead of using paid APIs, I created a MySQL database containing around 10 million records for the autocomplete functionality.
The problem is I need somewhere to host this database without spending money. I thought about two possible solutions:
- Setting up a virtual machine
- Converting my old smartphone (about 5 years old) into a server by rooting it
Which Linux distribution would work best for either approach? Should I go with the VM route or the phone server idea? Are there any other free alternatives I should consider for hosting this large database?
I’m pretty new to server management so any guidance would be helpful.
Google Cloud Platform and AWS both offer generous free tiers that could handle your requirements better than the options you mentioned. GCP gives you $300 credit for new accounts plus ongoing free usage limits, while AWS RDS free tier provides 20GB storage which might be tight but workable with proper data compression. I spent months wrestling with a similar autocomplete project and learned the hard way that 10 million records is overkill. Most effective autocomplete systems use frequency-based ranking and only store entries that have been searched for multiple times. You could probably trim that dataset to under 1 million records without losing meaningful functionality. If you absolutely must go the self-hosted route, DigitalOcean occasionally offers promotional credits for new users, and their $5 droplet would handle your database much better than a phone server. For Linux distributions, stick with Ubuntu Server LTS or Debian stable - you want something well-documented since you mentioned being new to server management. The smartphone idea might seem cost-effective but will become a maintenance nightmare once you factor in heat issues, limited RAM, and unreliable storage for database operations.
FreeTier cloud databases might not handle your 10 million records well due to connection and storage limits. I actually went through something similar last year and ended up using Oracle Cloud’s Always Free tier which gives you 20GB storage and 2 AMD compute instances permanently. The catch is their signup process can be tricky and they sometimes reject applications, but once you’re in it’s genuinely free forever. For your use case, I’d recommend Debian 11 if you go the VM route - it’s rock solid and uses less resources than Ubuntu. The phone server concept will fail hard with that dataset size, trust me on this one. Before committing to any hosting solution though, seriously consider implementing a search index like Elasticsearch or even just a simple trie data structure in memory. You probably don’t need all 10 million records accessible simultaneously for autocomplete. Most users only care about the first 5-10 suggestions anyway, so you could cache popular queries and load data dynamically. This approach would make almost any free hosting option viable.
honestly the phone server idea is kinda neat but gonna be a nightmare with 10mil records lol. id suggest checking out supabase free tier first - they give you 500mb which might work if you compress your data. also heroku postgres has free options tho limited. for vm route centos or debian would be solid choices, more lightweight than ubuntu sometimes.
Having dealt with similar database hosting challenges, I’d recommend looking into Railway or PlanetScale’s free tiers before considering the hardware routes you mentioned. Railway gives you 500 hours monthly which should be sufficient for development and testing phases, while PlanetScale offers decent storage limits for MySQL databases. The smartphone server approach sounds creative but will likely cause more headaches than it’s worth. Battery degradation, thermal throttling, and connectivity issues make it unreliable for production use. Plus, managing 10 million records on mobile hardware will be painfully slow. For your VM option, if you go that route, Ubuntu Server 22.04 LTS would be your safest bet due to its stability and extensive documentation. However, consider that you’ll need reliable internet infrastructure and proper backup solutions. Another angle worth exploring is optimizing your dataset first. Ten million records for autocomplete seems excessive - you might achieve better performance by preprocessing and reducing that dataset to maybe 100k-500k most relevant entries. This would open up more free hosting options and improve response times significantly.