keepcalculator

06 Numbers and Powers

How to Choose a Data Center Provider for Cloud & Colocation

Card No. 06·75402 · Updated · Standard formula, shown below

Choosing a data center provider for cloud and colocation workloads is a critical decision that affects performance, cost, reliability, and compliance. This guide walks you through the key technical, commercial, and operational factors to evaluate, so you can pick a provider that meets your current needs and scales with your future growth. Before you start, you’ll need a clear picture of your workloads, budget, and compliance requirements.

What You’ll Need

  • A list of your current and planned workloads (compute, storage, network)
  • Estimated power and cooling requirements per rack or per workload
  • Latency and geographic requirements (where are your users?)
  • Compliance or regulatory requirements (e.g., GDPR, HIPAA, SOC 2)
  • A budget range for monthly recurring costs and upfront setup
  • Your growth projections (how much will you need in 1, 3, 5 years?)

Step-by-Step

  1. Define your workload requirements. Start by categorizing your workloads: some may be better suited for cloud (bursty, elastic), others for colocation (steady, high-performance, or legacy). For each, note the required compute, storage, network bandwidth, and latency tolerance.
  2. Determine your power and cooling needs. Calculate the total power draw per rack (in kW) and the cooling density required. High-density workloads (e.g., AI/ML) need providers that offer liquid cooling or high-density cooling options. Verify that the provider can deliver the power density you need now and in the future.
  3. Evaluate network connectivity and carrier options. Check which carriers and internet exchanges are present at the facility. For cloud workloads, low-latency connections to major cloud providers (AWS, Azure, GCP) are crucial. Look for providers with direct cloud on-ramps or partner programs.
  4. Assess physical security and compliance certifications. Visit the facility if possible. Check for security measures like biometric access, CCTV, mantraps, and 24/7 guards. Verify current compliance certifications (SOC 2, ISO 27001, PCI DSS, etc.) and whether they meet your industry’s requirements.
  5. Review redundancy and uptime guarantees. Look for N+1 or 2N redundancy for power and cooling. Check the provider’s historical uptime record and their SLA (Service Level Agreement) for power, network, and cooling. Ensure the SLA includes financial penalties for downtime.
  6. Consider location and disaster recovery. Choose a location that balances latency to your users with disaster risk. Avoid putting all your infrastructure in one region. For colocation, consider a provider with multiple facilities so you can set up replication or failover.
  7. Compare pricing models and contract terms. Understand how you’re billed: per rack, per kW, per circuit, or a mix. Cloud providers typically charge per usage (compute hours, storage GB), while colocation is a monthly recurring fee. Look for hidden costs like cross-connects, remote hands, and power overage fees. Negotiate contract length and volume discounts.
  8. Test the provider’s support and operational processes. Ask about remote hands services, monitoring, and incident response. Check if they offer 24/7 support and how quickly they respond to tickets. Review their maintenance windows and how they communicate planned downtime.
  9. Check for scalability and future-proofing. Can the provider accommodate your growth in terms of space, power, and network? Are they investing in upgrades (e.g., higher power density, new cloud on-ramps)? Avoid providers that are at capacity or lack a clear expansion roadmap.
  10. Read the contract carefully. Look for exit clauses, data portability, and what happens if the provider is acquired or goes out of business. Ensure you can get your data and equipment out without excessive fees.

Common Mistakes

  • Underestimating power requirements. Many organizations calculate power based on nameplate ratings, which are higher than actual draw. Use actual measured power or realistic estimates to avoid paying for unused capacity.
  • Ignoring network latency. Choosing a facility far from your users or cloud providers can add significant latency. Always test real network paths before committing.
  • Overlooking cross-connect costs. Connecting to carriers or cloud providers often requires cross-connects that cost extra per month. These can add up quickly.
  • Focusing only on price. The cheapest provider may have poor uptime, bad support, or hidden fees. Consider total cost of ownership (TCO) including downtime costs.
  • Not planning for disaster recovery. Putting all your infrastructure in one facility without a DR plan is risky. Even with redundancy, a regional outage can take you down.

Tips

  • Start with a pilot. If possible, run a small workload or a few racks in the facility before migrating everything. This gives you real-world experience with their support and operations.
  • Use a hybrid approach. Combine cloud for bursty workloads and colocation for steady, predictable ones. This can optimize cost and performance.
  • Negotiate power overage rates. Some providers charge high rates for power above your contracted amount. Negotiate a reasonable overage rate or a buffer.
  • Check the facility’s power source. Ask about the utility feed and whether they have on-site generators and fuel storage. Understand the runtime of backup generators.
  • Ask for references. Talk to existing customers about their experience, especially regarding support and billing.

Troubleshooting

If you run into issues during the selection or migration, here’s what to do:

  • Provider doesn’t meet your power needs. If a facility can’t provide the density you need, look for a provider that offers high-density zones or liquid cooling. Don’t compromise on power—it will limit your growth.
  • Latency is higher than expected. Test with real applications and different carriers. Sometimes a different carrier or a dedicated connection can reduce latency. If not, consider a different location.
  • Costs are over budget. Review your usage and look for ways to right-size your footprint. Consider moving some workloads to cloud to reduce colocation space. Renegotiate your contract—many providers are flexible.
  • Compliance gaps. If the provider lacks a certification you need, ask if they can provide a letter of attestation or if they are in the process of obtaining it. If not, you may need to find a different provider.
  • Support is unresponsive. Raise the issue with your account manager. If it persists, consider having a remote hands provider on standby or look for a new provider.

Frequently Asked Questions

What is the difference between cloud and colocation?

Cloud services are provided on-demand over the internet, with you paying for usage and the provider managing the hardware. Colocation means you rent space, power, and cooling in a data center and provide your own servers and storage. Cloud offers flexibility and scalability, while colocation gives you full control over hardware and often lower costs for steady workloads.

How much power do I need per rack?

Typical racks use 3-10 kW, but high-density workloads like AI/ML can require 20-50 kW or more. Calculate your actual power draw based on your equipment’s specifications and usage patterns. Always leave some headroom for growth.

What uptime SLA should I look for?

Look for providers offering at least 99.9% uptime (about 8.7 hours of downtime per year). For mission-critical workloads, aim for 99.99% (about 52 minutes per year). Ensure the SLA includes financial credits if the uptime is not met.

Can I connect to multiple cloud providers from a colocation facility?

Yes, many colocation providers offer direct connections to major cloud providers via cross-connects or partner programs. This reduces latency and data transfer costs compared to using the public internet.

What is a cross-connect and why does it cost extra?

A cross-connect is a physical cable that connects your equipment to another provider’s equipment within the same data center (e.g., a carrier or cloud provider). It costs extra because it requires physical installation and ongoing maintenance. Prices vary by provider and bandwidth.

How long does it take to migrate to a data center?

It depends on the scale. A small migration (a few racks) can take weeks, while a large enterprise migration can take months. Plan for time to install equipment, configure networks, and test. Many providers offer migration services to help.

Related

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.

More cards in this drawer

The whole Numbers and Powers drawer →