Colocation Facility Constraints for Infrastructure Migration¶

Scope¶

This file covers colocation facility constraints that must be addressed when planning hardware deployments, migrations, or expansions in third-party data centers (Equinix, NTT, Digital Realty, CyrusOne, QTS, and similar providers). These constraints are frequently overlooked during architecture design and can introduce weeks or months of delay if not identified early. Applies to any project involving physical hardware in a colocation facility, whether greenfield deployment, migration from one colo to another, or expansion within an existing facility. For facility lifecycle management, see general/facility-lifecycle.md. For physical server assessment, see general/physical-server-scope.md.

Checklist¶

Why This Matters¶

Colocation facility constraints are the most common source of "invisible" project delays because they involve physical-world lead times that cannot be compressed with additional engineering effort. A cross-connect that takes 10 business days to provision cannot be accelerated by working overtime. A facility that has no available power circuits requires either waiting for infrastructure upgrades (months) or finding alternative space. During migrations, the dual-stack period -- where both old and new environments must run simultaneously -- doubles power and cooling requirements temporarily, and many colocation contracts do not include provisions for temporary capacity increases. Hardware delivery to a colocation facility is not the same as delivery to an office: loading docks may be shared across multiple tenants, freight elevators have scheduled availability, and cage access may require escort by facility staff. Remote hands costs can escalate quickly during intensive migration periods if on-site staff presence is not planned, as even simple tasks like power-cycling a server or reseating a cable incur per-incident fees. Failing to account for these constraints during the architecture phase results in migration timelines that are technically correct but operationally impossible.

Common Decisions (ADR Triggers)¶

Same-facility migration vs facility change -- Migrating within the same colocation facility (rack-to-rack or cage-to-cage) eliminates WAN latency concerns and allows direct fiber cross-connects between old and new environments, but is constrained by facility capacity. Changing facilities provides an opportunity to renegotiate contracts and adopt better infrastructure but introduces inter-facility connectivity complexity and longer dual-stack periods. Same-facility is strongly preferred when capacity exists.
Remote hands vs dedicated on-site staff during migration -- Remote hands (provider-supplied technicians) are cost-effective for occasional tasks ($75-$200/hour, billed per incident) but introduce response time delays (15 min to 4 hours) and may lack familiarity with your specific hardware. Dedicated on-site staff (your own or contracted) cost more in travel and labor but provide immediate response and continuity. For migrations spanning more than 2-3 days of active rack work, dedicated on-site presence is typically more cost-effective and reliable.
Direct carrier cross-connects vs meet-me room patching -- Direct cross-connects to a carrier's cage are simpler and lower-latency but lock you to that carrier. Meet-me room patching allows access to multiple carriers and IX connections from a single demarcation point but adds a patch panel hop. For redundancy, most deployments use cross-connects to at least two carriers via the MMR, with IX peering for high-traffic public-facing services.
Full rack vs partial rack (caged colo vs shared colo) -- Full racks in a private cage provide maximum physical security, flexible power configurations, and unrestricted access but carry higher costs ($800-$2500+/month per rack depending on market and power). Shared or partial-rack colocation is cheaper but limits power density, restricts physical access scheduling, and may impose neighbor noise or vibration concerns. Migrations are significantly easier with full-rack private cage access.
Power redundancy: 2N vs N+1 vs single feed -- 2N (fully redundant, dual independent power paths) provides the highest availability and is standard for production workloads in Tier III+ facilities. N+1 (single path with redundant components) is cheaper and sufficient for non-critical or dev/staging workloads. Single feed is only appropriate for equipment that tolerates brief outages. During migration, ensure new racks match or exceed the redundancy level of existing infrastructure.
Contract structure: committed vs burst capacity -- Long-term committed capacity (1-3 year terms) provides the best per-unit pricing but requires accurate capacity forecasting. Burst or flex capacity provisions allow temporary increases during migrations but at premium pricing (20-50% above committed rates). Negotiating burst provisions into the base contract before migration begins is significantly cheaper than requesting ad-hoc capacity increases.

Rack Space and Power Planning¶

Consideration	Typical Values	Migration Impact
Standard rack height	42U-48U	Plan for 30-36U usable after PDU, cable management, airflow
Power per rack (standard)	4-8 kW	Verify sufficient for modern high-density servers (1-2 kW each)
Power per rack (high density)	10-20+ kW	May require liquid cooling or rear-door heat exchangers
Single circuit capacity	20A or 30A at 208V single-phase	20A circuit = ~3.3 kW usable at 80% derating
Three-phase power	30A at 208V three-phase	~8.6 kW usable at 80% derating, more efficient for dense racks
Dual-stack overhead	2x normal power/cooling	Budget for full overlap period, negotiate temporary circuits
Rack monthly cost	$800-$2500+	Varies heavily by market (NoVA/Ashburn cheapest, NYC/SF premium)

Power Planning for Migration¶

During a migration, both old and new infrastructure must run concurrently. Plan power accordingly:

Inventory existing power draw -- Measure actual draw per rack via PDU metering, not rated capacity
Estimate new hardware power requirements -- Use vendor power calculators (Dell, HPE, Lenovo) for accurate per-server estimates
Calculate total concurrent draw -- Old + new must not exceed available circuits
Request temporary circuits from the provider if needed -- lead time is typically 2-4 weeks for additional circuits within an existing cage
Plan power surrender schedule -- As old racks are decommissioned, notify the provider to release circuits and reduce billing

Cross-Connect and Connectivity Lead Times¶

Connection Type	Typical Lead Time	Notes
Intra-facility copper cross-connect	3-7 business days	Within same building, MMR to cage
Intra-facility fiber cross-connect	5-15 business days	Single-mode LC/SC, verify MMR port availability first
Inter-facility dark fiber	15-45 business days	Between buildings on same campus, requires strand availability
New carrier circuit (metro ethernet)	30-60 business days	Carrier must have presence in facility or provision last-mile
New carrier circuit (wavelength/DWDM)	45-90 business days	Long-haul or inter-market connectivity
IX port provisioning	5-20 business days	Requires IX membership approval, port availability
Temporary migration cross-connect	5-10 business days	Request removal scheduling at time of order to avoid orphaned connections

Key point: Always order cross-connects as early as possible in the migration timeline. They are on the critical path and cannot be parallelized with other preparation work.

Facility Access and Security Procedures¶

Access Models by Provider¶

Provider Type	Access Model	Typical Requirements
Premium (Equinix IBX, Digital Realty)	Badge access, 24/7, self-service for authorized personnel	Photo ID, background check, NDA, access request 24h in advance for new personnel
Standard (regional providers)	Escorted access during business hours, badge after-hours	Escort fee ($50-$100/visit), advance booking required
Shared colo	Scheduled access windows only	Reservation required, limited to 2-4 hour windows

Access Planning for Migration¶

Build the authorized personnel list early -- adding new personnel can take 3-5 business days at some facilities
Book loading dock time for hardware delivery -- 24-48 hours advance notice is standard
Coordinate with other tenants if shared loading dock -- avoid conflicts during heavy delivery periods
Plan escort coverage for contractors -- facility staff escorts may need to be pre-booked
Document after-hours access procedures -- migration cutovers often happen on weekends or overnight

Hardware Delivery and Staging Logistics¶

Vendor Ships Hardware
        |
        v
Loading Dock Receiving ───── Requires: dock scheduling (24-48h advance)
        |                           freight elevator booking
        v                           receiving contact on authorized list
Staging Area Unpacking ───── Requires: staging space (may be limited/shared)
        |                           inventory verification
        v                           asset tagging
Cage/Suite Transport ─────── Requires: cage access (badge + escort if needed)
        |                           cart or dolly availability
        v
Rack Mounting ────────────── Requires: rail kits, cage nuts, power cables
        |                           PDU port allocation
        v                           cable management plan
Cabling and Patching ─────── Requires: patch panel ports
        |                           labeled cables (both ends)
        v                           cross-connect completion
Power-On and Burn-In ─────── Requires: power circuit activation
        |                           IPMI/BMC network configuration
        v                           remote access verification
Production Ready

Common pitfalls: - Hardware delivered to wrong dock or without advance scheduling sits in limbo - Rail kits and power cables not ordered with servers (different SKUs, often forgotten) - PDU outlets already fully allocated in existing racks - Cage nuts and mounting hardware missing from rack kit - Staging area too small for palletized server shipments

Decommission Procedures¶

Data destruction -- All storage media must be sanitized per NIST 800-88 (Clear, Purge, or Destroy) before hardware leaves the facility. Maintain chain-of-custody documentation with serial numbers, sanitization method, and verification signatures.
Cable removal -- Remove all patch cables and label/bundle for reuse or disposal. Request cross-connect removal from the provider (typically 3-5 business day lead time, billed as a disconnect fee).
Power release -- Notify the provider of circuits to be released. Power circuits on committed contracts may continue to incur charges until the contract term ends or is amended.
Rack teardown -- Remove all equipment, shelves, and cable management. If returning rack space to the provider, schedule a walkthrough for condition verification.
Asset disposition -- Arrange for pickup by ITAD (IT Asset Disposition) vendor, internal logistics, or provider disposal service. Schedule loading dock and freight elevator for outbound shipment.
Contract amendment -- Submit formal request to reduce committed capacity. Lead time for contract amendments is typically 30-60 days, and reductions may not take effect until the next billing cycle.

Remote Hands Service Tiers¶

Service Level	Response Time	Typical Cost	Use Case
Standard	2-4 hours	$75-$125/hour	Non-urgent: cable audits, visual inspections
Priority	30-60 minutes	$125-$175/hour	Moderate urgency: power cycling, cable reseating
Emergency	15-30 minutes	$175-$250/hour	Critical: hardware failure response, emergency power
Dedicated technician	On-site, continuous	$500-$1500/day	Extended migration work, multi-day deployments

Tips: - Pre-authorize specific remote hands tasks with the provider to avoid approval delays during incidents - Maintain a detailed rack diagram and cable schedule that remote hands technicians can follow without verbal guidance - Some providers offer "smart hands" (higher-skill technicians for OS-level or network tasks) at premium rates - Budget for 20-40 remote hands hours during a typical multi-rack migration

Colocation Provider Comparison Factors¶

When evaluating or comparing colocation providers, especially during a facility migration:

Factor	What to Assess
Carrier neutrality	Number of carriers in MMR, IX presence, ease of provisioning new circuits
Power density options	Maximum kW per rack, availability of high-density (15-20+ kW) racks
Contract flexibility	Minimum commitment terms, burst provisions, early termination penalties
Geographic proximity	Distance to users, other facilities, and cloud on-ramps
Compliance certifications	SOC 2 Type II, ISO 27001, PCI DSS, HIPAA, FedRAMP
Interconnection ecosystem	Cloud direct connect (AWS, Azure, GCP) availability within facility
Expansion capacity	Ability to add racks/cages without changing facilities
Remote hands quality	Response time SLAs, technician skill level, cost structure

Reference Resources¶

Uptime Institute Tier Standards: uptimeinstitute.com -- Tier I-IV facility classification, redundancy requirements, and availability targets (Tier III: 99.982%, Tier IV: 99.995%)
ASHRAE TC 9.9: Data center cooling and environmental guidelines -- recommended temperature ranges (18-27C intake), humidity guidelines, and allowable envelopes for server hardware
NIST SP 800-88 Rev. 1: Guidelines for Media Sanitization -- required reading for decommission planning, defines Clear/Purge/Destroy methods by media type
TIA-942: Data center infrastructure standard -- rack layout, cable pathway, and grounding/bonding requirements for colocation environments
PeeringDB: peeringdb.com -- verify IX presence and peering participants at specific facilities before signing contracts