Cloud strategy for an operator: public, private, hybrid — a decision frame

When the fork appears

The operator decides on infrastructure strategy. Public cloud (AWS / Azure / GCP / regional providers) promises elasticity and managed services. Private cloud (on-prem or managed private) provides control and data residency. Hybrid complicates operations but gives flexibility.

In UZ an additional variable is local clouds (Beeline Cloud, Uztelecom Cloud) with regulatory positioning but a limited managed services portfolio.

Per-workload frame criteria

Data residency. If the regulator requires PII in country — public cloud is usable only if the vendor has a local region. Most hyperscalers do not have one in UZ — a disqualifier for customer-facing data.

Latency. Real-time decisioning, network functions, IVR require <50ms latency. Public cloud is far — a disqualifier. Local cloud or edge — a must.

Workload variability. If workload is spiky (campaigns in seasonal peaks) — public cloud elasticity wins. If steady (biller) — private is cheaper.

Managed services dependency. If you actively use managed databases, ML services, serverless — public cloud is cheaper in TCO. If everything is custom — the benefit is smaller.

Team skill set. Public cloud requires cloud-native skills (DevOps, SRE, IaC). Private — traditional infra skills. Hire/retrain — a separate bottleneck.

Vendor lock-in. Public cloud creates managed-services lock-in. Private creates hardware/licence lock-in.

Cost predictability. Private — predictable CapEx + run cost. Public — variable, may surprise as it grows.

Per workload split

Customer-facing apps & decisioning. Local cloud or edge for latency.

CRM & analytics for marketing. Public cloud (if data residency allows) — managed services give velocity.

Biller & rating. Private — critical operations, low risk tolerance, data residency.

Network functions (vRAN, packet core). Edge / private with telecom-grade specifics.

ML training. Public cloud for elasticity; model serving — where low latency is required.

DR / backup. Multi-region on public as safety.

Where decisions usually go wrong

“Everything on public cloud”. A year later bill shock, data residency violations, regulator scrutiny.

“Everything private”. Three years later infrastructure is stale, business velocity drops, talent leaves for more modern shops.

Hybrid without discipline. Each team makes its own choice, nobody knows what is where — operational chaos.

Lift-and-shift. VMs moved to public cloud without re-architecture. TCO higher than on-prem was.

Vendor lock-in without an exit story. Five years later the vendor raises prices, exit is impossible.

When which approach

All-in public: small operators (<1M customers), no significant data residency, willing to skip infra capability building.

All-in private (on-prem or managed): strict regulatory environment, large operator with strong infra team, capex willingness.

Hybrid (most common for mid-size): customer-facing on local/edge, analytics on public, biller on private, ML mixed.

What to discuss at the committee

Data residency requirements per workload — regulator clarification.

Per-workload 5-year TCO for each scenario.

Skill-set audit and hire/retrain plan for the chosen path.

Vendor lock-in mitigation strategy.

Exit costs from each vendor.

Regulator dialogue around cloud usage — some regulators require notification.

How SamaraliSoft helps

Cloud Strategy Decision Workshop — 4-6 weeks. Workload inventory, per-workload assessment, TCO modelling, vendor evaluation, regulatory mapping, target architecture with migration roadmap.

Related

• /en/architecture/telecom-around-core-architecture/ — wrapping the core
• /en/insights/telecom-vendor-management/ — vendor management
• /en/insights/telecom-platform-tco/ — TCO
• /en/decisions/telecom-build-vs-buy-cdp/ — CDP decision