Colocation Checklist: What to Ask Before Signing

Choosing a colocation provider is a long-term commitment. Most contracts run 12-36 months, and moving hardware between data centres is disruptive and expensive. Before signing, make sure you have asked the right questions. This checklist covers everything from power and cooling to contract terms, drawn from our experience placing GPU clusters and enterprise servers across Indian data centres.

Power and Cooling

Power is the single most critical factor for hardware-heavy deployments, especially GPU servers.

What is the maximum power density per rack (kW)? Most Indian Tier 3 facilities offer 4-8 kW per rack as standard. GPU deployments need 15-30 kW per rack. Confirm the facility can deliver what you need.
How is power metered and billed? Some facilities charge a flat rate per rack (with a kW cap), others meter actual consumption. Metered billing is fairer for variable workloads; flat rate is simpler for budgeting.
What is the power redundancy architecture? Look for 2N (fully redundant) or N+1 (one spare unit) UPS and generator systems. Ask for the UPS runtime on battery and the generator switchover time.
Are dual power feeds available per rack? Dual PDU feeds from independent power chains protect against single-feed failure. Essential for production deployments.
What is the PUE (Power Usage Effectiveness)? PUE measures total facility power divided by IT equipment power. A PUE of 1.4-1.6 is typical in India; newer facilities achieve 1.2-1.3. Lower PUE means less money spent on cooling overhead.
What cooling technology is used? Raised-floor CRAC/CRAH, in-row cooling, hot-aisle containment, rear-door heat exchangers, or direct liquid cooling. For dense GPU racks, ask whether the facility supports rear-door heat exchangers or liquid cooling loops.
Can the facility support liquid cooling? If you plan to deploy liquid-cooled GPU servers (which are becoming more common for H100/H200 and next-gen GPUs), verify that the facility has chilled water or coolant distribution infrastructure.

Network and Connectivity

Is the facility carrier-neutral? A carrier-neutral data centre allows you to choose from multiple ISPs and carriers, ensuring competitive pricing and redundancy.
Which ISPs and carriers are present? Verify that your preferred ISPs (Airtel, Jio, Tata Communications, BSNL, ACT Fibernet, etc.) have a point of presence in the facility.
Is internet exchange peering available? Facilities connected to NIXI (National Internet Exchange of India) or private peering exchanges offer lower-latency, higher-bandwidth connectivity.
What are the cross-connect costs? Budget for INR 3,000-10,000 per month per cross-connect, plus one-time installation fees.
Is dark fibre available for site-to-site connectivity? If you need dedicated links to other facilities or your office, verify that dark fibre providers service the building.
What bandwidth options are available? Check availability of 1G, 10G, and 100G Ethernet ports. For GPU clusters, you may need 100GbE uplinks to the internet and a separate InfiniBand fabric within your cage.
Are cloud on-ramps available? Direct Connect (AWS), ExpressRoute (Azure), and Cloud Interconnect (Google Cloud) require presence in the same facility or a connected one.

Physical Security and Access

What are the physical security measures? Look for multi-factor authentication (biometric + card), mantrap entry, 24/7 CCTV with 90+ day retention, and on-site security personnel.
What are the access hours? 24/7 access is standard for most facilities. Confirm there are no restrictions on weekend or holiday access.
How quickly can visitors be authorised? For emergency maintenance, you may need to add a technician to the access list on short notice. Ask about the authorisation process and timeline.
Is there a secure storage or staging area? Useful for storing spare hardware, unpacking new equipment, and staging before rack installation.

Remote Hands and Support

Is remote hands available 24/7? Some facilities only offer remote hands during business hours. For production GPU clusters that may need emergency intervention at 3 AM, 24/7 availability is essential.
What is the remote hands response time SLA? Look for 15-30 minute response time guarantees for standard requests.
What is the skill level of remote hands staff? Can they handle BIOS configuration, drive replacement, cable re-patching, and console access, or only basic power cycling?
What is the pricing model? Per-incident, per-hour, or included hours per month. Budget for at least 4-8 hours of remote hands per month for a multi-server deployment.

SLA and Reliability

What is the uptime SLA? Tier 3 facilities should guarantee 99.982% or better. Ask for the actual uptime track record over the past 12-24 months, not just the contractual SLA.
What are the SLA penalty terms? Service credits for downtime are standard, but verify the calculation method and maximum credit amount. Some SLAs cap credits at one month’s fees, not very meaningful for a major outage.
What is the planned maintenance schedule? Tier 3 facilities should be concurrently maintainable (no downtime for routine maintenance), but confirm this explicitly.
What is the disaster recovery plan? Ask about fire suppression (gas-based, not water), flood protection, earthquake resilience, and insurance coverage.

Contract and Commercial Terms

What is the minimum contract term? Typically 12-36 months. Shorter terms are available but at higher monthly rates.
What are the setup fees? One-time charges for rack preparation, power provisioning, and initial cross-connects.
How does pricing escalate? Are there annual price increases? Are they capped? CPI-linked or fixed percentage?
What is the exit process? How much notice is required to terminate? Are there early termination penalties? What happens to your cross-connects and IP addresses?
Is there a right to audit? For compliance purposes, you may need the ability to audit the facility’s security, power, and cooling infrastructure.

Compliance and Certifications

What certifications does the facility hold? Look for ISO 27001 (information security), SOC 2 Type II (service controls), and Uptime Institute Tier certification. In India, some facilities also hold PCI-DSS certification for financial services workloads.
Does the facility meet RBI/SEBI data localisation requirements? If you handle financial data, verify that the facility’s compliance posture meets regulatory requirements.
Is there support for DPDP Act compliance? The Digital Personal Data Protection Act, 2023, requires organisations to demonstrate control over personal data processing. Colocation provides a clear physical control narrative.

GPU-Specific Considerations

If you are deploying GPU servers, add these to your checklist:

Can the facility support 15-30 kW per rack? Standard 4-6 kW racks are insufficient for GPU servers.
Is there sufficient cooling for GPU density? GPU servers generate concentrated heat loads that require enhanced airflow or liquid cooling.
Are there weight restrictions per rack? A rack of GPU servers can weigh 500-1,000+ kg. Verify floor loading capacity.
Is there space for network switches? GPU clusters need top-of-rack switches (Ethernet and possibly InfiniBand). Confirm rack space allocation for networking equipment.
Can the facility support InfiniBand cabling? InfiniBand uses different cable types and bend radius requirements than standard Ethernet. Verify that the cable management infrastructure accommodates this.

Our Recommended Process

Shortlist 3-4 facilities based on location, carrier availability, and stated power density
Schedule site visits to inspect the data hall, power infrastructure, and cooling systems firsthand
Request a technical proposal with specific rack power, cooling, and connectivity details for your deployment
Review the contract with attention to SLA penalties, exit terms, and price escalation
Negotiate: most commercial terms are negotiable, especially for multi-rack or multi-year commitments

rawcompute.in has established relationships with Tier 3 colocation providers across Mumbai, Chennai, Pune, Hyderabad, and Bengaluru. We can help you evaluate facilities, negotiate terms, and coordinate hardware delivery and installation. Contact us for assistance with your colocation selection process.