Cloud Cost Optimization in 2020: Escaping the Hyperscaler Trap Post-Schrems II

The CFO just walked into your office. Or, more likely given the current state of the world this July, they pinged you on Slack with a screenshot of the AWS billing dashboard. It’s red. It’s ugly. And with the Norwegian Krone (NOK) taking a beating against the Dollar this year, your infrastructure costs have effectively risen 15% without you spinning up a single new instance.

To make matters worse, the Court of Justice of the European Union (CJEU) dropped a nuclear bomb on the industry four days ago (July 16, 2020) by invalidating the Privacy Shield framework in the Schrems II ruling. Suddenly, hosting personal data on US-owned clouds—even in their EU regions—is a legal minefield regarding GDPR compliance.

As a CTO, you have two problems: your bill is too high, and your legal standing is shaky. We are going to solve both by returning to engineering fundamentals. It is time to stop paying for "managed magic" and start optimizing raw compute.

1. The "Egress Tax" and Hidden IOPS Costs

The biggest lie in modern cloud computing is the low hourly rate. You see a cheap instance price, but you ignore the bandwidth costs. Hyperscalers charge distinct fees for data leaving their network. If you run a media-heavy site or an API utilized by partners across Europe, this egress fee can rival your compute costs.

Pro Tip: Audit your invoice for "Data Transfer Out." If it exceeds 20% of your total bill, you are on the wrong platform. At CoolVDS, we allocate generous bandwidth pools with our NVMe instances because we peer directly at NIX (Norwegian Internet Exchange). We don't penalize you for being successful.

2. Technical Rightsizing: Stop Over-Provisioning

Developers often request resources based on peak theoretical load rather than average utilization. In a containerized world (Docker/Kubernetes), this leads to "bin packing" inefficiencies where you pay for reserved CPU cycles that sit idle.

Before you scale out, scale up and tune. A single well-tuned VPS often outperforms a complex cluster of micro-instances due to the lack of network overhead between services.

The Silent Killer: RAM Swapping

If your application hits the swap file, your performance falls off a cliff. However, over-allocating RAM is expensive. The solution is aggressive tuning of your data stores.

Optimizing MySQL 8.0 / MariaDB 10.4 for VPS

The default configurations for MySQL are often archaic. If you are running a database on a VPS with 8GB RAM, but your innodb_buffer_pool_size is set to the default 128MB, you are wasting IOPS and RAM simultaneously. The database will constantly read from disk (slow, consumes I/O credits) instead of memory.

Here is a baseline configuration for a standard 8GB RAM instance running a LEMP stack. Add this to your /etc/my.cnf:

[mysqld]
# Use 60-70% of RAM for DB if it's a dedicated DB server
# Use 30-40% if running on the same web server
innodb_buffer_pool_size = 3G

# Reduce disk I/O by increasing log file size
innodb_log_file_size = 512M

# Crucial for writing data safely but efficiently on SSD/NVMe
innodb_flush_method = O_DIRECT
innodb_io_capacity = 1000

# Connection handling
max_connections = 150
open_files_limit = 65535

3. Nginx Caching: The CPU Saver

Every time PHP-FPM or Python generates a page, you burn CPU cycles. If you are serving content that doesn't change every second (like a blog post or a product description), you must use FastCGI caching. This allows Nginx to serve the page directly from RAM/Disk without waking up the application backend.

This single change can allow a $20/month CoolVDS instance to handle traffic that would crush a $100/month setup without caching.

# inside nginx.conf
fastcgi_cache_path /var/run/nginx-cache levels=1:2 keys_zone=WORDPRESS:100m inactive=60m;
fastcgi_cache_key "$scheme$request_method$host$request_uri";

# inside server block
server {
    set $skip_cache 0;

    # POST requests and urls with a query string should always go to PHP
    if ($request_method = POST) { set $skip_cache 1; }
    if ($query_string != "") { set $skip_cache 1; }

    location ~ \.php$ {
        fastcgi_pass unix:/var/run/php/php7.4-fpm.sock;
        fastcgi_cache WORDPRESS;
        fastcgi_cache_valid 200 60m;
        fastcgi_cache_bypass $skip_cache;
        fastcgi_no_cache $skip_cache;
        
        include fastcgi_params;
    }
}

4. Benchmarking Storage: NVMe Matters

In 2020, spinning rust (HDD) is obsolete for primary workloads. Even standard SSDs are becoming a bottleneck for high-traffic databases. The metric that matters is IOPS (Input/Output Operations Per Second) and Latency.

Many providers cap your IOPS based on disk size (e.g., 3 IOPS per GB). This forces you to buy storage you don't need just to get the speed you do need. This is a "capacity tax."

Let's look at a simple fio benchmark test you can run to verify your current provider's claims vs. CoolVDS NVMe storage:

# Install fio
apt-get install fio -y

# Run a random read/write test simulating a database load
fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 \
    --name=test --filename=test --bs=4k --iodepth=64 --size=1G \
    --readwrite=randrw --rwmixread=75

On a standard cloud "General Purpose" SSD, you might see 300-600 IOPS. On CoolVDS local NVMe storage, we frequently clock over 10,000 IOPS on similar workloads. Higher IOPS means your queries finish faster, your CPU waits less (low I/O wait), and you can downgrade to fewer cores while maintaining performance.

5. The Legal "Cost": Schrems II and Datatilsynet

We cannot ignore the legal reality. The invalidation of the Privacy Shield means that relying on Standard Contractual Clauses (SCCs) to transfer data to US providers is now under intense scrutiny. Datatilsynet is clear about data controller responsibilities.

Moving your workload to a Norwegian provider like CoolVDS isn't just about latency (though pinging Oslo in 2ms is nice); it's about risk mitigation. Hosting data physically in Norway, under Norwegian jurisdiction, simplifies your GDPR compliance posture immediately. The cost of a legal audit or a GDPR fine far outweighs the cost of migration.

Conclusion: Efficiency is the New Growth

The era of throwing credit cards at AWS to solve performance problems is over. The economic climate of 2020 demands precision. By rightsizing your instances, implementing aggressive caching, and moving to high-IOPS NVMe storage that doesn't charge for bandwidth, you can cut your monthly burn rate significantly.

Stop paying the "Cloud Tax." Audit your infrastructure today.

Is your cloud bill bleeding revenue? Deploy a high-performance, GDPR-ready NVMe instance on CoolVDS today and experience the difference of local Norwegian infrastructure.