Choosing a data center for corporate infrastructure is a decision whose consequences you will live with for years. Moving equipment from one facility to another is a labour-intensive, risky operation: it requires a maintenance window, planned downtime, and significant resources. Getting the initial choice wrong is expensive.

Here are the 7 criteria that actually drive the decision.

1. Reliability tier

The first and most important parameter is the reliability level defined by the Uptime Institute classification. Four tiers describe what a facility can withstand without losing availability:

  • Tier I — basic level, single path for power and cooling. Planned maintenance requires a shutdown.
  • Tier II — redundant components added, but still a single distribution path.
  • Tier III — concurrently maintainable: any component can be serviced without stopping IT. Uptime ~99.982% (1.6 hours downtime per year).
  • Tier IV — fault tolerant: any single unplanned failure is survived automatically. Uptime ~99.995% (26 minutes downtime per year).

Practical rule: Tier III suits most enterprise workloads. Tier IV is needed where even brief unplanned downtime causes real financial or operational damage: financial systems, payment processing, peak-season e-commerce, healthcare, government infrastructure.

Important: a tier rating counts only with an Uptime Institute certification, not self-declaration. Always ask for the certificate.

2. Location and connectivity

Geographic location affects three things: latency to your users, compliance with data localisation requirements, and the logistics of physical access for your team.

  • Latency. For most business applications, 10–20 ms is a comfortable threshold. If your users are in Almaty and the data center is in Astana, calculate the latency. For latency-sensitive applications, choose a facility in the same city.
  • Data localisation. Under Kazakhstani law, personal data of RK citizens must be physically stored in Kazakhstan. A data center in Kazakhstan is the only way to guarantee compliance with this requirement.
  • Physical access. Your team will need to reach the facility for non-standard work. A location 500 km away significantly increases operational costs.

Connectivity is another critical parameter. Look for:

  • Number of independent carriers connected to the facility (ideally 3 or more).
  • Physically separate cable routes (important protection against a single cable cut).
  • Presence of an internet exchange point (IX) or direct peering agreements.
  • Latency to key destinations: Moscow, Almaty, Astana, Tashkent, international hubs.

3. Power capacity and density

Insufficient power capacity is one of the most common causes of problems as infrastructure grows. Ask about:

Power per rack. Standard is 5–10 kW per rack. High-performance computing, GPU servers, and AI infrastructure require 20–50+ kW per rack. Confirm the facility can provide the right density in your specific placement zone.

Power headroom. How much electrical capacity does the facility have available over the next 3–5 years? A facility running at 90% load cannot accommodate your expanding infrastructure.

Power redundancy scheme. Tier III uses an N+1 scheme with a single distribution path. Tier IV uses a 2N scheme with independent paths. Clarify how redundancy is actually implemented: how many utility feeds, what UPS systems, diesel generators with what fuel reserve.

PUE (Power Usage Effectiveness). The ratio of total facility power consumption to IT equipment consumption. PUE 1.3–1.5 is good; PUE below 1.3 is excellent. A high PUE means a significant portion of your electricity bill goes to infrastructure rather than your servers.

4. Physical and information security

The facility where your company’s data is stored must provide multi-layer protection:

Perimeter and physical access:

  • 24/7 on-site security, video surveillance with recording.
  • Access control system (cards/biometrics).
  • Two-factor entry to the server hall.
  • Full visitor logging — also a compliance requirement.
  • Dedicated cage or locked cabinet available for sensitive data.

Fire suppression:

  • Gas suppression (FM-200, Novec 1230, inert gas) — not water.
  • Early smoke detection (VESDA).
  • Fire compartments with independent suppression per zone.

Information security:

  • ISO 27001 — the minimum standard for a serious data center.
  • Optional: SOC 2, PCI DSS (for financial data).

5. Support and SLA

In a crisis, support quality matters more than contract language. Evaluate:

Remote hands. Who will physically work with your equipment per your instructions? How fast will they respond? What hours is the service available? What is the cost?

SLA response times. What does the data center promise for a power loss? A connectivity issue? A cooling failure? An SLA without financial penalties for breach is not a real SLA.

NOC (Network Operations Center). Is there a 24/7 on-duty team monitoring the infrastructure and responding before you call?

Incident history. Ask for a report on major incidents over the past 2–3 years. This shows how the data center behaves in abnormal situations, not just in marketing materials.

6. Scalability

Infrastructure decisions are made for years. Confirm that the data center has room to grow:

  • Available rack space or ability to add racks without a relocation.
  • Power headroom for new workloads.
  • Option to move to a dedicated cage or a full room as you scale.
  • Flexible contract terms: short-term start with renewal options.

Also assess the ecosystem. Are there cloud providers, carriers, or partners in the facility that enable convenient direct peering? An ecosystem of partners reduces response time and the cost of connectivity.

7. Regulatory compliance

For certain industries, compliance is not optional — it is a licensing condition:

Data localisation (critical for Kazakhstan). The RK Personal Data Law requires storing data of Kazakhstani citizens in Kazakhstan. The data center must be able to provide documentary confirmation of the physical location of servers.

Financial sector. NBK and ARRFR requirements for information security at banks and MFIs specify placement in certified facilities.

Government and quasi-government organisations. MCRIAP requirements for state information systems call for placement in data centers that have passed the relevant accreditation.

International standards. If the business works with European clients — ISO 27001 and GDPR compatibility. For payment systems — PCI DSS.

Checklist before signing a contract

Before signing:

  • Uptime Institute certificate (if the facility claims a Tier rating)
  • ISO 27001 or other security certifications
  • Actual power redundancy scheme (not marketing copy)
  • Number of independent network carriers
  • Available power per rack in your zone
  • SLA with financial penalties for breach
  • Remote hands terms and pricing
  • Physical address of the facility in Kazakhstan (for compliance)
  • Option to audit and visit the facility
  • Contract termination and equipment relocation terms

Frequently asked questions

How do I choose a data center for a small business?

For a small company with 1–5 servers, the key criteria are: location (latency and access), reliability of at least Tier III, transparent rack-space and traffic pricing, and responsive support. Tier IV is justified if your applications are mission-critical and downtime is costly.

What matters more — location or reliability tier?

Both criteria matter, and for different workloads the priority shifts. For latency-sensitive applications, location matters more. For critical systems where downtime is unacceptable, reliability tier matters more. Ideally, the facility satisfies both requirements.

How do I verify a claimed Tier rating?

Ask the data center for the Uptime Institute certificate — a document with a QR code verifiable at uptime.com. Statements like “meets Tier III requirements” without a certificate are marketing, not fact.

Should I visit the facility before choosing?

Strongly recommended. A site tour lets you assess the real state of the infrastructure, ask engineers questions directly, see how server halls are organised, and confirm the data center has its own NOC. Any serious facility is always open to such visits.

What is PUE and why does it matter?

PUE (Power Usage Effectiveness) is the ratio of total facility energy consumption to IT equipment consumption. PUE = 2.0 means that for every watt consumed by servers, another watt is spent on infrastructure (cooling, lighting, UPS). PUE 1.3 is significantly more efficient. This directly affects your electricity bill in colocation arrangements with per-kWh billing.

What is the difference between Tier III and Tier IV data centers?

Tier III guarantees concurrent maintainability: any component can be serviced without stopping IT, eliminating planned downtime as a risk. Tier IV adds fault tolerance: any single unplanned component failure is survived automatically without interruption. Tier IV uptime is ~99.995% vs ~99.982% for Tier III — roughly 26 minutes vs 1.6 hours of downtime per year.

Ready to evaluate Akashi Data Center as a home for your infrastructure? Contact us — we will arrange a facility tour and calculate the cost of colocation for your requirements.