Data-Center Classification by Type of Operation
Type Primary purpose / activity Latency (RTT) Aggregate Max fiber distance to Technical highlights
bandwidth backbone / POP
Enterprise / Internal workloads (ERP, databases, 10–50 ms 1–10 Gbps (dual < 50 km Two independent carriers; focus on security and
Corporate intranet). Typically owned and operated (regional) 10G typical) reliability; single-tenant; planned maintenance
by a single company. tolerated.
Colocation Hosts third-party equipment with shared 1–20 ms 10–100 Gbps per < 20 km Carrier-neutral; multiple fiber routes; redundant
(Multi-tenant) power and cooling infrastructure. tenant; ≥400 Gbps network fabrics; SLA ≥99.98%.
total backbone
Cloud / Hyperscale Large-scale public cloud (AWS, Azure, < 1 ms 100 Gbps – > 1 < 5 km Private DWDM backbone; 2N redundancy;
Google). Designed for massive intra-campus; < Tbps (internal availability zones within 50 km; multi-Tbps fabric
east-west traffic and self-healing 5 ms inter-zone backbone) capacity.
operation.
Edge / Micro-Edge Small modular facilities near end-users 1–5 ms to user 1–40 Gbps < 5 km 1–5 racks typical; ultra-low latency; multi-carrier local
for low-latency services (IoT, gaming, access; strong QoS enforcement.
CDN).
Carrier Hotel / IXP Neutral interconnection hubs between 0.1–1 ms 400 Gbps – > 10 < 1 km Core peering point; extremely high fiber density;
ISPs and cloud operators (DE-CIX, Tbps >99.99% uptime; hundreds of cross-connects.
AMS-IX, LINX).
Industrial / Supports SCADA, energy, or 5–20 ms (local) 1–10 Gbps < 50 km (or < 10 km Emphasis on isolation; may use redundant
On-Premises / manufacturing systems; focused on local loop) microwave backhaul; autonomous local operations.
Modular Private local reliability.
Practical Interpretation
1. Latency (RTT) — Round-trip delay between data center and end-user or interconnection point:
• Under 1 ms: ultra-critical (trading, AI inference, real-time gaming).
• 1–5 ms: edge / cloud workloads with low jitter.
• 5–10 ms: regional clouds and most enterprise apps.
• 20–50 ms: backups, analytics, batch workloads.
2. Aggregate Bandwidth — Total capacity across uplinks:
• Hyperscale: 100G / 400G / 800G DWDM per fiber pair.
• Colocation: at least 2×10G redundant uplinks.
• Edge: 10–40G with strict QoS.
3. Distance to Backbone / POP
• About 0.5 ms latency per 10 km of fiber.
• Keep under 20 km for sub-5 ms services.
• Beyond 50 km behaves as a regional, not metro, data center.