Colocation

Managed vs Unmanaged Colocation: Choosing the Right Service Level for Your Infrastructure

Every colocation buyer faces a fundamental decision before signing a contract: how much operational responsibility do you want to retain, and how much do you want the provider to handle? This question defines the boundary between managed and unmanaged colocation -- two service models that look identical from the outside (same racks, same power, same cooling) but differ dramatically in who does what after the hardware is installed.

The decision is not purely financial. It involves staffing, compliance, response time requirements, workload complexity, and the strategic question of whether infrastructure operations is a core competency your organization needs to build or a function best outsourced. This guide breaks down both models in detail, compares their costs and trade-offs, and provides a decision framework for colocation buyers evaluating providers in the UAE and globally.

What Unmanaged Colocation Actually Includes

Unmanaged colocation -- sometimes called "bare" or "wholesale" colocation -- provides the physical infrastructure layer and nothing above it. The provider delivers the environment; the customer operates everything inside it.

What the Provider Delivers

• Physical space: Rack, cage, or suite with locked cabinets and raised floor or slab construction. Physical dimensions, weight limits, and airflow orientation are specified in the contract.
• Power: Metered utility power at contracted capacity (measured in kW or kVA), with redundancy per the SLA (N+1, 2N, or concurrent maintainable). Power distribution units (PDUs) deliver power to the rack level. The provider maintains the electrical infrastructure up to the PDU output; the customer handles power cords and internal distribution.
• Cooling: Environmental control to maintain ASHRAE-specified temperature and humidity ranges. In the UAE, this means mechanical cooling running continuously -- there is no free cooling window at 45-50 degrees Celsius ambient temperatures.
• Network connectivity: Cross-connects to carrier meet-me rooms and internet exchange points. The provider maintains the physical fiber and copper infrastructure. The customer procures their own IP transit, peering, or private network services from carriers available in the facility.
• Physical security: Multi-layer access control (biometrics, badges, mantrap entries), CCTV surveillance, and 24/7 security staff. The provider controls who enters the facility; the customer manages their own access lists.

What the Customer Is Responsible For

Everything above the physical infrastructure layer falls to the customer:

• Hardware procurement and installation: Shipping equipment to the facility, racking and cabling servers, switches, and storage devices. This includes labeling, documentation, and physical verification of installations.
• Operating system and firmware: OS installation, patching, kernel updates, BIOS and firmware updates for all deployed hardware. The customer schedules and executes maintenance windows.
• Monitoring: The customer deploys and operates their own monitoring stack (Nagios, Zabbix, Datadog, Prometheus, or similar). The provider monitors facility-level metrics (power draw, ambient temperature, humidity) but not individual servers.
• Incident response: When a server fails, the customer must dispatch their own technician to the facility or use the provider's remote hands service (a separate, pay-per-use offering). Response time depends on the customer's staffing and geography.
• Backup and disaster recovery: The customer designs and operates their own backup strategy, replication, and disaster recovery architecture.
• Compliance: While the provider maintains facility-level certifications (ISO 27001, SOC 2 for the physical plant), the customer is responsible for compliance at the application, data, and access layers. This includes audit evidence, access logs, encryption, and regulatory reporting.

Remote Hands: The Middle Ground

Most unmanaged colocation providers offer remote hands as an add-on service. Remote hands means the provider's on-site technicians perform physical tasks at the customer's direction: rebooting a server, swapping a failed drive, re-seating a cable, photographing a status panel. The customer tells the technician exactly what to do; the technician executes without making independent decisions about the customer's infrastructure.

Remote hands is a reactive service. The technician does not monitor, diagnose, or proactively maintain the customer's equipment. If a disk fails and the monitoring system does not alert anyone, the failed disk sits until the customer notices. This is the defining limitation of the unmanaged model: the provider sees the facility, but not the customer's workload.

What Managed Colocation Adds

Managed colocation layers operational services on top of the physical infrastructure. The provider takes responsibility for tasks that would otherwise require the customer to maintain their own operations team.

Core Managed Services

• Hardware deployment and lifecycle: The provider receives, inventories, racks, cables, and commissions new hardware. They manage hardware warranty claims, coordinate RMA replacements with manufacturers, and handle end-of-life equipment decommissioning. For GPU deployments, this includes specialized tasks like installing liquid cooling manifolds, connecting InfiniBand cables, and verifying GPU thermal sensor readings post-installation.
• 24/7 monitoring and alerting: The provider monitors all customer equipment at the hardware level (CPU temperature, fan speed, disk health, memory errors, power supply status) and optionally at the OS level (service availability, CPU/memory/disk utilization, network throughput). Alerts are triaged by the provider's NOC before escalation to the customer, reducing alert fatigue.
• OS and firmware management: The provider schedules and applies operating system patches, security updates, BIOS updates, and firmware upgrades according to an agreed change management process. This is particularly valuable for large fleets where patch management across hundreds of servers is a full-time job.
• Proactive incident response: When monitoring detects a hardware failure, the provider initiates repair or replacement without waiting for the customer to notice and request action. A failed power supply is swapped from on-site spares inventory. A degraded RAID array is rebuilt. A server with memory errors is taken offline and reseated. The customer receives a notification after the issue is resolved, not a request to authorize action.
• Backup and recovery operations: The provider runs backup jobs on an agreed schedule, monitors backup success and failure, rotates media, and tests restores periodically. The customer defines the backup policy (what to back up, how often, how long to retain); the provider executes it.
• Compliance support: The provider assists with compliance audits by providing access logs, change management records, patch histories, and incident reports in formats required by auditors. For UAE TDRA compliance and data residency requirements, managed providers typically maintain the documentation framework that satisfies regulatory auditors.

Extended Managed Services

Some providers offer deeper management layers beyond hardware and OS:

• Network management: Firewall rule management, VPN configuration, load balancer tuning, DDoS mitigation, and traffic engineering. The provider operates the customer's network stack as a service.
• Database administration: Database performance tuning, query optimization, replication management, backup verification, and failover testing. Available for common platforms (PostgreSQL, MySQL, MongoDB, Redis).
• Application monitoring: APM (Application Performance Monitoring) integration that tracks application-level metrics such as response time, error rates, throughput, and user experience scores. This crosses into managed hosting territory and is typically priced separately.
• Security operations: Log aggregation, SIEM (Security Information and Event Management) integration, vulnerability scanning, and incident investigation. Some providers operate a dedicated SOC (Security Operations Center) as an add-on service.

Cost Comparison: The Full Picture

Comparing managed and unmanaged colocation purely on the provider's monthly invoice misses most of the cost equation. The real comparison is total cost of ownership (TCO), which includes the customer's internal costs.

Cost Component	Unmanaged	Managed
Facility fee (per kW/month)	$150-250	$220-400
Remote hands (estimated monthly)	$500-2,000	Included
Internal NOC staff (allocated)	$5,000-15,000	$0-2,000
Monitoring tools (licenses)	$500-3,000	Included
Travel for site visits	$1,000-5,000	$0-500
Compliance documentation	Internal labor	Included
Incident response time	Variable (30 min - 8 hrs)	15 min - 1 hr SLA

The Crossover Point

For most organizations, the cost crossover between managed and unmanaged colocation occurs between 30 and 100 kW of deployed power. Below 30 kW (roughly 10-15 racks at moderate density), the customer cannot justify a full-time operations engineer, and the managed service premium is less than the cost of ad-hoc remote hands plus monitoring tools plus occasional travel. Above 100 kW (roughly 30-50+ racks), the customer can hire a dedicated on-site engineer whose fully loaded cost ($120,000-180,000/year in the UAE market) is distributed across enough equipment to be cheaper per kW than the managed service premium.

These examples illustrate the scale-dependent nature of the decision. Small deployments almost always cost less with managed services when all internal costs are included. Large deployments benefit from the economies of dedicated internal staff.

Service Level Agreements: Where the Models Diverge

The SLA in an unmanaged colocation contract covers the facility: uptime of power, cooling, and network connectivity. A typical colocation SLA guarantees 99.99% or 99.999% infrastructure availability. If the facility loses power (not caused by the customer's equipment), the provider credits a portion of the monthly fee.

Managed colocation SLAs extend further. In addition to facility uptime, the provider commits to:

• Hardware replacement time: A guaranteed window for replacing failed components from on-site spares (typically 4 hours for standard components, 24 hours for specialized parts like GPU modules).
• Incident response time: How quickly the provider's NOC acknowledges and begins working on an alert (typically 15 minutes for critical alerts, 1 hour for warnings, 4 hours for informational).
• Patch deployment time: For critical security patches, the SLA may require deployment within a specified window (e.g., 72 hours for CVSS 9.0+ vulnerabilities).
• Backup verification: Regular backup restore tests with documented results provided to the customer.
• Reporting frequency: Monthly or quarterly operational reports covering uptime, incidents, patches applied, capacity trends, and recommendations.

The SLA structure directly affects financial exposure. An unmanaged colocation customer bears the full cost of downtime caused by their own operational failures (a missed firmware update that causes a crash, a failed disk that was not replaced because monitoring did not catch it). A managed colocation customer can hold the provider accountable for operational failures within the provider's scope, potentially recovering costs through SLA credits.

Workload-Specific Considerations

Bitcoin Mining and ASIC Hosting

ASIC mining operations have specific characteristics that influence the managed vs. unmanaged decision. Mining hardware is relatively simple from an OS perspective -- most ASIC miners run embedded firmware with minimal configuration. The primary operational tasks are physical: monitoring hash rate per unit, identifying and replacing failed units, managing power distribution across circuits, and handling the high heat output of dense mining deployments.

For miners, the decision often comes down to fleet size. A miner with 10-50 units in a single facility usually benefits from managed hosting, where the provider handles the physical monitoring and replacement cycle. A large mining operation with 500+ units in multiple facilities often employs its own operations team, using unmanaged colocation with remote monitoring software that provides per-unit hash rate visibility. The hosting contract should clearly define what constitutes a "failed unit" and the replacement process regardless of which model is chosen.

GPU and AI Infrastructure

GPU colocation introduces complexity that favors managed services for most customers. A high-density GPU rack running at 30-60 kW presents thermal challenges that require active management. Liquid cooling systems need regular coolant quality checks, flow rate monitoring, and leak detection. InfiniBand fabrics require firmware coordination across switches and NICs. NVIDIA driver and CUDA toolkit updates must be sequenced carefully to avoid breaking running workloads.

A managed GPU colocation provider with AI hosting expertise brings specialized knowledge that most enterprise IT teams lack. They know which NVIDIA driver versions are stable for production, how to configure MIG partitioning for multi-tenant environments, and how to diagnose GPU memory errors versus training instability. This expertise commands a premium, but it prevents the costly mistakes that generalist teams make when first deploying GPU infrastructure.

Enterprise and Financial Services

Enterprise customers in regulated industries (banking, healthcare, government) often require managed colocation for compliance reasons. Auditors expect to see documented change management, patch histories, and incident response records maintained by qualified personnel. Producing this documentation from an internal team managing equipment in a remote colocation facility is possible but labor-intensive. A managed provider delivers audit-ready reports as part of the service, significantly reducing the internal compliance burden.

The UAE regulatory environment adds additional considerations. TDRA data classification requirements, Central Bank CBUAE technology risk guidelines, and Dubai International Financial Centre (DIFC) data protection regulations all specify operational controls around data handling. A managed colocation provider familiar with UAE regulatory requirements can implement these controls as part of their standard operating procedures, while a customer using unmanaged colocation must build and document the compliance framework independently.

Hybrid Models: The Practical Middle Ground

In practice, many colocation deployments use a hybrid approach that combines elements of both models. Common hybrid configurations include:

• Unmanaged colocation + managed network: The customer operates their servers but outsources network management (firewalls, load balancers, DDoS mitigation) to the provider. This is common for organizations with strong systems teams but limited network engineering depth.
• Unmanaged colocation + enhanced remote hands: The provider's remote hands team is pre-trained on the customer's standard procedures (documented runbooks for common failure scenarios). This gives the customer faster, more accurate remote hands response without the full cost of managed services.
• Managed colocation + customer-retained control plane: The provider manages hardware and OS, but the customer retains control of all application deployment, configuration management (Ansible, Puppet, Chef), and monitoring dashboards. This gives the customer visibility and control without the operational burden of hardware management.
• Tiered management by criticality: Production systems under managed services (maximum SLA protection), development and staging environments under unmanaged colocation (lower cost, more flexibility for the customer's engineering team to experiment).

The best colocation providers are flexible about service boundaries. Rather than offering a fixed "managed" or "unmanaged" package, they work with customers to define a scope of services that matches the customer's internal capabilities and gaps. The contract negotiation should explicitly document which tasks belong to the provider and which belong to the customer, with no ambiguity in the handoff points.

Decision Framework: How to Choose

Use the following criteria to determine which colocation model fits your deployment:

Choose Unmanaged Colocation When:

• You have an existing operations team with data center experience.
• Your deployment exceeds 50-100 kW and justifies dedicated on-site staff.
• You require maximum control over every aspect of the infrastructure (custom OS builds, non-standard configurations, experimental hardware).
• Your workload involves proprietary or classified systems where third-party access to the OS layer is restricted by policy or regulation.
• You operate in multiple facilities and have standardized your operations tooling to work across all of them.

Choose Managed Colocation When:

• Your deployment is under 50 kW and you cannot justify a dedicated operations engineer.
• Your team lacks specialized hardware skills (GPU infrastructure, liquid cooling, HPC workload management).
• You need guaranteed response times backed by SLA credits, not best-effort internal response.
• Compliance requirements demand documented operational controls maintained by a certified provider.
• Your primary business is not infrastructure operations, and you want to focus engineering resources on application development rather than hardware management.
• You are deploying in a geography (such as the UAE) where you do not have local staff and time zone coverage is a challenge.

Consider a Hybrid Model When:

• You have strong skills in some areas (application, security) but gaps in others (hardware, network).
• You are scaling from a small deployment to a large one and want to transition from managed to unmanaged as your team grows.
• You have mixed workloads with different criticality levels and different management requirements.
• You want to retain strategic control (architecture, vendor selection, capacity planning) while outsourcing operational execution.

Evaluating Managed Colocation Providers

Not all managed colocation offerings are equal. When evaluating providers, assess the following:

• Staff qualifications: What certifications do the on-site operations team hold? For GPU infrastructure, look for NVIDIA DGX Certified Administrators. For general data center operations, look for CDCP (Certified Data Centre Professional) and vendor-specific hardware certifications.
• Spares inventory: Does the provider maintain on-site spares for common failure components (drives, power supplies, DIMMs, fans, network cables)? What is the spares policy for specialized hardware (GPU modules, InfiniBand HCAs, liquid cooling components)? A managed provider without adequate spares inventory cannot deliver on their replacement SLA.
• Change management process: How does the provider handle changes to the customer's systems? Is there a documented change advisory board (CAB) process? Are changes logged, tested, and reversible? The quality of the change management process is a leading indicator of operational maturity.
• Escalation paths: What happens when the on-site team cannot resolve an issue? Is there an engineering escalation tier with deeper expertise? Can the customer escalate directly, or must all communication go through the NOC?
• Reporting and visibility: Does the customer get real-time dashboard access to monitoring data, or only periodic reports? The best managed providers give customers full visibility into the same monitoring systems the NOC uses, so the customer can verify SLA compliance independently.
• Exit provisions: If the customer wants to transition from managed to unmanaged (or to a different provider), what is the process? Is there a knowledge transfer period? Are there early termination penalties specific to the managed services component?

The UAE Market Context

The managed vs. unmanaged decision in the UAE has several local factors that differ from markets like the United States or Europe.

Limited local talent pool. Data center operations engineers with deep hardware experience are scarce in the UAE market. International recruitment is common but adds visa processing, relocation, and retention costs. This scarcity makes managed colocation more attractive for organizations that cannot compete for the limited available talent.

Regulatory momentum. The UAE is rapidly expanding its data protection and technology risk regulations. Organizations deploying data center infrastructure need operational controls that satisfy regulators. Managed providers that specialize in the UAE market typically maintain compliance frameworks aligned with TDRA, CBUAE, ADGM, and DIFC requirements, saving customers the cost of building and maintaining these frameworks internally.

Climate-specific operations. Data center operations in the UAE involve considerations that are absent in temperate markets: extreme ambient temperatures that stress cooling systems year-round, sand and dust infiltration that requires specialized filtration maintenance, and humidity control challenges in coastal cities. A managed provider with UAE-specific operational experience handles these environmental factors as routine, while an operations team imported from a cooler climate faces a learning curve that risks equipment damage.

Time zone advantage for global workloads. The UAE sits at UTC+4, which provides overlapping business hours with Europe, Africa, South Asia, and East Asia. For organizations distributing workloads globally, a managed colocation facility in the UAE provides operational coverage during hours when a US-based team is offline, complementing rather than duplicating coverage.