I have been hosting several websites (WordPress and others) on AWS for many years and one my favorite ways to manage Wordpress websites on my servers has been using Plesk.
But as it occurred to me that the cost of hosting and yearly pricing of Plesk itself has been going up over the years, I needed to find a better way to cost-effectively manage my websites.
Now, I am not a Unix/Linux expert and have not managed servers. And hiring server admins will simply cost me way more than Plesk (especially since they have good support to deal with any server issues that need human intervention). So, I decided to check with Cloude.ai about my confusion and just explore if this was an idea worth exploring.
So, I asked Claude.ai this:
I have a “Plesk Obsidian for AlmaLinux – WordPress and Website Hosting” on AWS for which I pay for EC2 instance m5a.large and then I pay for Plesk license separately yearly around more than 1000 dollars. I host about 15 wordpress websites on this server. I feel like just to host 15-20 WordPress website that are not too heavy, this is too much price to pay. Is there a better way to host my websites on AWS without paying this much money and have the server be more optimized to take some additional workloads by using the resources for other AI related work or app developments etc.?
In response, Claude.ai started out this way:
You’re absolutely right that you’re overpaying for what you’re using. Let me break down better alternatives:
Current Costs Analysis
- EC2 m5a.large: $744/year)
- Plesk license: ~$1,000/year
- Total: ~$1,744/year for 15 WordPress sites
This is roughly $116/site/year, which is excessive for basic WordPress hosting.
Better AWS Alternatives
Option 1: AWS Lightsail (Best for simplicity)
- Cost: $3.50-$5/month per WordPress instance
- For 15 sites: Use 3-4 Lightsail instances ($10-20/month each) = ~$40-80/month ($480-960/year)
- Includes managed WordPress, automatic backups, SSL certificates
- Savings: ~$800-1,300/year
Option 2: Smaller EC2 + Free Control Panel
- EC2 t3a.medium or t3a.small: $12-24/month
- Use free alternatives to Plesk:
- Webinoly (free, lightweight NGINX stack)
- WordOps (free, optimized for WordPress)
- CyberPanel (free, with OpenLiteSpeed)
- HestiaCP (free, fork of VestaCP)
- Cost: ~$15-30/month ($180-360/year)
- Savings: ~$1,400/year
Option 3: Hybrid Approach (Maximize m5a.large usage)
Keep your m5a.large but optimize it:
For WordPress sites:
- Switch to free control panel (save $1,000/year on Plesk)
- Use Docker containers for better resource isolation
- Implement caching (Redis/Memcached) for better performance
Utilize remaining resources for:
- AI/ML workloads: Run Ollama, LLaMA models, or small AI agents
- Development environments: Docker containers for testing
- Database server: PostgreSQL/MySQL for other projects
- CI/CD pipelines: GitLab Runner, Jenkins
- Media processing: Video transcoding, image optimization
- Cost: Just EC2 (~$744/year)
- Savings: $1,000/year, PLUS you get compute for AI/development
My Recommendation
Start with Option 2 or 3:
- Migrate to t3a.medium ($24/month) + WordOps/CyberPanel (free)
- Handles 15-20 WordPress sites easily with proper caching
- Save ~$1,400/year
- Simple migration from Plesk
- If you want to explore AI/development, keep the m5a.large and:
- Install CyberPanel or HestiaCP (free)
- Use Docker to containerize WordPress sites (better resource management)
- Allocate remaining 8GB RAM and 2 vCPUs for:
- Local AI models (Ollama with smaller models)
- Development environments
- Testing/staging servers
This was exciting to see that I could switch out of Plesk without impact and that there were multiple pathways to achieve the goal. So, I asked Claude to go with option 3 of keeping my server but just get rid of Plesk and save money I was spending on just Plesk and it’s addons. So, I asked it to give me a detailed step-by-step plan to achieve this goal and then it produced a very nice guide like this:
What Happened When I Asked For Developer’s Help
I have known a few “AWS experts” and a few developers whom I thought to show this plan and see if they could help me achieve this in a reasonable cost and time so that I don’t have to learn Linux server management and I could use my time for something more productive (like writing, experimenting with new AI features etc. which I love doing). So, I shared this document with these experts and asked their opinion or interesting in helping me out with this.
They took 2-3 weeks to just get back to me, and when they did, they send me a quote for $4000+ and about 45 days of time to get this done (basically, they had to follow the guidelines I had shared). It made me upset and disappointed at the same time although I knew instantly that if this needs to be done, I need to do it myself because I was not going to spend $4000+ on a server switching exercise! Plus, I didn’t have 45-50 days of time where I pay them, and here I pay Amazon and Plesk!
My Decision: If You Want It Done, Do It Yourself!
Days later, on one Saturday afternoon, I decided to try implementing this document and follow the server migration instructions in the documents one-step at a time and see how far I could go. So, I started.
But wait, I thought, let me check if m5a.large is the best server to continue to have? Or should I get a better server if it comes at the same price or cheaper? So, I asked Claude again:
Should I continue with m5a.large EC2 instance itself or do you recommend a better instance for the same or better load and same or less cost?
In response, it did a bit of thinking and then said this:
Great question! Let me analyze better options for your workload. The m5a.large might not be the most cost-effective choice anymore.
Better Alternatives at Similar or Lower Cost
Recommended: t3a.xlarge ⭐
Specs:
- 4 vCPUs (double the m5a.large)
- 16GB RAM (double the m5a.large)
- Cost: $1,452/year)
- Burstable performance with unlimited mode
Why This is Better:
- More resources for roughly the same price
- T3 instances get CPU credits – perfect for WordPress (bursty traffic)
- Double the RAM = more room for AI models (can run 7B-13B models comfortably)
- Better value: $0.0752/hour vs m5a.large at $0.086/hour
- Handles 25-30 WordPress sites easily
- Much better for AI/development work
Budget Option: t3a.large 💰
Specs:
- 2 vCPUs (same as m5a.large)
- 8GB RAM (same as m5a.large)
- Cost: $744/year)
- Saves ~$40/month vs m5a.large
Why Consider This:
- Half the cost of m5a.large
- Same RAM, same vCPUs
- T3 burstable performance is better for WordPress workloads
- Handles 15-20 WordPress sites fine
- Still room for smaller AI models (3B-7B parameters)
Performance Option: t3.xlarge (if you need guaranteed baseline)
Specs:
- 4 vCPUs
- 16GB RAM
- Cost: $1,740/year)
- Better network performance than t3a
- Intel vs AMD processors
Direct Comparison
| Instance | vCPUs | RAM | Cost/Month | Cost/Year | Best For |
| m5a.large | 2 | 8GB | $62 | $744 | Legacy choice |
| t3a.large | 2 | 8GB | $62 | $744 | Budget (15-20 sites) |
| t3a.xlarge ⭐ | 4 | 16GB | $121 | $1,452 | Best value |
| m5a.xlarge | 4 | 16GB | $124 | $1,488 | Sustained high CPU |
| t3.xlarge | 4 | 16GB | $145 | $1,740 | Intel, better network |
My Recommendation: t3a.xlarge
Reasons:
- Perfect for WordPress – Burstable CPU is ideal for web traffic patterns
- Better AI capability – 16GB RAM lets you run larger models (Mistral 7B, Llama2 13B)
- Growth room – Can handle 25-30 sites vs your current 15
- Cost effective – Best performance per dollar
- Unlimited mode – Set CPU credits to unlimited for consistent performance
Understanding T3 “Burstable” Performance
Don’t let “burstable” scare you:
- T3 instances accumulate CPU credits when idle
- WordPress sites are idle 90%+ of the time
- You can enable Unlimited mode = never throttled (small extra charge only if you consistently max out CPU)
- For 15 WordPress sites, you’ll rarely burn through credits
CPU Baseline:
- t3a.xlarge baseline: 40% of 4 vCPUs = 1.6 vCPUs sustained
- This is plenty for 15-20 WordPress sites
- You get burst to 100% when needed
So, based on this recommendations, I did my own research a little bit and decided to go with the recommended t3a.xlarge server as it was a better option all the way without adding to the cost.
Should I stay with AlmaLinux vs Ubuntu on my new EC2 Instance?
Like everything else, I asked this question too in the context of this server migration project and although it gave below details about each one of these and their pros and cons, the verdict was that I should stay with AlmaLinux9 for a stable production quality operating system.
AlmaLinux 9 (RHEL-based) – Recommended for You
Pros:
- 10-year support lifecycle (until 2032)
- Rock solid stability – enterprise-grade
- SELinux enabled by default – better security
- Perfect for CyberPanel – officially supported and tested
- Slower update cycle = fewer breaking changes
- Better for production servers that you don’t want to touch often
- YUM/DNF package manager – very stable
- Compatible with RHEL ecosystem – tons of enterprise software
Cons:
- Slightly older packages (prioritizes stability over cutting-edge)
- Smaller community than Ubuntu (but still large)
- Some newer software requires manual installation
- Documentation sometimes RHEL-focused
Best for:
- Production web servers
- Long-term stability
- Enterprise workloads
- Set-it-and-forget-it servers
Ubuntu 24.04 LTS (Debian-based)
Pros:
- Massive community – more tutorials, Stack Overflow answers
- APT package manager – huge software repository
- Newer packages by default
- Better for Docker – more Docker tutorials use Ubuntu
- Easier AI/ML setup – most guides assume Ubuntu
- 5-year support (can extend to 10 with Ubuntu Pro – free for personal use)
- Snap packages – easy installation of modern software
- More familiar to most developers
Cons:
- More frequent updates = potential for issues
- Less focused on enterprise stability
- Security updates can occasionally break things
- Shorter default support window (5 vs 10 years)
Best for:
- Development environments
- AI/ML workloads
- Rapid deployment of new software
- When you need cutting-edge packages
Next Important Question: What Control Panel to Use on New Server?
Claude did its research and recommended that I should be using CyberPanel, a free control panel, good for AlamaLinux and a good fit for my situation. Well, here is what it said:
CyberPanel is a modern, open-source web hosting control panel built on OpenLiteSpeed, offering a streamlined alternative to traditional panels like cPanel or Plesk. When paired with AlmaLinux, a stable and enterprise-grade RHEL-based distribution, it creates an ideal hosting environment that’s both cost-effective and reliable. The combination particularly shines for users seeking a free control panel solution without sacrificing performance or security features.
The key benefit of running CyberPanel on AlmaLinux is the blend of speed and stability. OpenLiteSpeed’s lightweight architecture delivers impressive performance for WordPress and other web applications, while AlmaLinux provides the rock-solid foundation and long-term support that enterprises expect. This pairing is especially attractive for hosting providers and system administrators who want to reduce licensing costs while maintaining a professional-grade hosting stack with features like automated SSL, one-click installations, and email management all accessible through an intuitive web interface.
Migration Action – Start to Finish of The Big Server Migration
So, I started on the journey with the Phase 1, then Phase 2, and then Phase 3 as recommended in the guideline with the t3a.xlarge server as the only major change and continued step by step. I am not going to publish all the steps, and back-and-forth with Claude wherever there were some technical challenges and how I got them resolved with the help of Claude again. But in a nutshell, here is what happened overall:
Phase 1: Setting Up the New Server
Phase 2: First Website Migration – This very blog was the first one to be migrated!
Phase 3: Scaling to ALL the remaining websites (WordPress and Non-WordPress)
Phase 4: The “Ghost in the Machine” — Caching Conflicts (Some websites had intermittent load issues due to W3 Total Cache’s cached data from old server that got carried over into the new server. But Claude helped getting those resolved finally with ease.
The AI Advantage: How Claude Made It Possible
Having AI is like having an all-knowing personal assistant 24/7 at your disposal and that’s what gave me the confidence to get the work started in the first place!I am on a Pro plan with Claude made it all so easy from start to finish! In fact this is the only reason why I was able to complete in less than 15 working hours for what I had been proposed a $4000+ with 45-50 days of working timeframe by the traditional experts! AI plan with Claude costs me less than $20 a month. But It helped me do a $4000+ job with ease to my satisfaction in less than a full day.
Anywhere I got stuck, anywhere I got an error or anywhere a website didn’t behave the way it’s supposed to after migration, all I had to do was the copy/paste the error into my Claude window or take a screenshot of the error and upload it. Rest of all the diagnosis, remediation steps and guidance was provided by Claude like a sincere assistance and it kept on sensibly providing commands, scripts and instructions with an intention to help me resolve the issue! It is an amazing experience working with AI and proves the point of how one can be 10X more productive by knowing how to use AI for solving business problems or even personal productivity problems!
Final Thoughts: The Future of Technical Work
This project convinced me we’re entering an era where anyone willing to learn can accomplish expert-level technical work with AI assistance.
I’m not a server administrator by trade or by education. But I am an IT professional with general knowledge about LINUX and know how to work with FileZilla, Putty and basic SSH commands. With that much of knowledge and with Claude’s help, I:
- Migrated 15 production websites
- Debugged complex caching issues
- Optimized server performance
- Saved $1,200/year for years to come, starting now!
The barrier isn’t technical knowledge anymore—it’s willingness to engage with AI as a learning partner rather than just a command generator.
If you’re running WordPress sites on expensive managed hosting, consider this path. The savings are real, the performance improvements are measurable, and with AI assistance, it’s more accessible than ever.
Have you attempted a similar migration? What challenges did you face? Share your experience in the comments!
Disclaimer: Server management carries risks. Always maintain backups and test thoroughly before migrating production sites. AI assistants provide guidance but cannot guarantee results—verify commands before execution.