Cloud Storage vs. Dedicated SAN: A 2010 Reality Check for Norwegian IT

It is the classic dilemma for any CTO in Oslo right now: do you sign another three-year lease on a NetApp SAN for your rack at the Digiplex facility, or do you trust your data to this rapidly expanding concept of "Cloud Storage"? If you have ever spent a Friday night rebuilding a degraded RAID-5 array while watching the iowait spike on your primary database, you know that storage performance is the single biggest bottleneck in modern infrastructure.

But the landscape in 2010 is shifting faster than most procurement cycles can handle. With the maturity of Xen hypervisors and the arrival of reliable Enterprise SSDs (like the Intel X25 series), the argument for massive, dedicated hardware arrays is thinning. However, moving to a Virtual Private Server (VPS) environment introduces new risks—specifically regarding "noisy neighbors" and I/O contention. Here is how we navigate the transition without sacrificing uptime or compliance.

The Latency Trap: Why Spindles Are Slowing You Down

Most hosting providers in Norway are still stacking SATA drives in software RAID. It is cheap, dense, and utterly incapable of handling high-concurrency transactional databases. When your MySQL Innodb_buffer_pool fills up and hits the disk, those 7,200 RPM drives incur a seek penalty that kills page load times.

In our tests comparing standard SATA VPS setups against 15k SAS and emerging SSD-backed tiers, the difference is not just measurable; it is transformative.

Benchmark: Random Read IOPS (4K Block Size)

If you are running a high-traffic e-commerce site targeting the Norwegian market, that latency difference translates directly to conversion rates. A user in Tromsø pinging a server in Amsterdam already deals with network hops; adding 15ms of disk latency on the backend is unacceptable.

Tuning Linux for Virtualized Storage

Moving to a virtualized environment requires tuning the Linux kernel to respect that it is not running on bare metal. The default I/O scheduler in RHEL 5 or Debian Lenny is usually CFQ (Completely Fair Queuing). On a dedicated server, CFQ is fine. On a VPS, where the hypervisor handles the physical disk sorting, CFQ adds unnecessary overhead.

We recommend switching your scheduler to deadline or noop inside your VM to reduce CPU steal time and let the host handle the heavy lifting.

Configuration Example:

# Check current scheduler
cat /sys/block/xvda/queue/scheduler
[cfq] deadline noop

# Change to deadline immediately
echo deadline > /sys/block/xvda/queue/scheduler

# Make it permanent in GRUB (menu.lst)
kernel /vmlinuz-2.6.18-194.el5 ro root=/dev/VolGroup00/LogVol00 elevator=deadline

Data Sovereignty: The Datatilsynet Factor

Technical performance is meaningless if you violate the law. The Norwegian Personal Data Act (Personopplysningsloven) places strict requirements on how personal data is handled. While the EU Safe Harbor framework theoretically allows data transfer to the US, the reality for Norwegian businesses is that keeping data within the EEA—and preferably within Norway—simplifies compliance drastically.

Pro Tip: Using US-based cloud giants (like Amazon EC2) for storage often introduces latency and legal grey areas regarding the Patriot Act. Hosting locally with a provider physically located at the NIX (Norwegian Internet Exchange) ensures your data stays under Norwegian jurisdiction and reaches local users via the shortest network path.

Why CoolVDS Bets on Hardware RAID-10

We do not believe in overselling storage. Many "cloud" providers oversell their disk space, betting that users won't fill their allocation. This leads to the "I/O Wait" death spiral during peak hours.

At CoolVDS, our architecture relies on Hardware RAID-10 arrays with battery-backed cache controllers. We utilize Xen HVM virtualization, which provides better isolation than OpenVZ containers. This ensures that a neighbor compiling a kernel does not consume your disk I/O. We are also aggressively rolling out SSD caching tiers, bridging the gap between mechanical capacity and flash speed.

The Verdict

The era of buying your own hard drives is ending. The Total Cost of Ownership (TCO) for maintaining a SAN, replacing failed disks, and paying for rack space in Oslo is simply too high compared to high-performance managed hosting. By choosing a provider that prioritizes low latency and local compliance, you gain the flexibility of the cloud without the performance penalty of cheap virtualization.

Ready to eliminate I/O bottlenecks? Stop watching iostat and start shipping code. Deploy a high-availability instance on CoolVDS today and experience the stability of true hardware isolation.