Geographically distributed infrastructure: why businesses need multiple points of presence

Dmitry Kukushkin, a technical writer at the cloud provider IT-GRAD, explained how geographically distributed infrastructure is structured, what tasks it solves, and how it helps businesses.

Dmitry Kukushkin, city — Almaty, technical writer at the cloud provider IT-GRAD

One Data Center — One Point of Failure

Typically, when deploying infrastructure, companies rent racks or cloud resources in a single data center. However, by 2026, this approach is already obsolete, especially for large businesses. Infrastructure downtime due to a provider-side failure or a cyberattack can cost millions of tenge.

The causes of incidents can vary. For example, even with power redundancy, a data center may sometimes fail to switch over without a brief interruption in the event of an emergency. Fire, flooding, human error during scheduled maintenance — all these scenarios are possible despite the high level of security in modern data centers.

The fundamental problem with a single site is that any incident renders all of the company’s services unavailable. Businesses have no choice but to wait and wonder if they are even operational.

At the moment, it is practically impossible to assess losses — both reputational and financial.

What is Geographically Distributed Infrastructure

Geographically distributed infrastructure is the simultaneous placement of resources across multiple sites, integrated into a single managed environment.

The key word here is “single.” Geographically distributed infrastructure implies that data is synchronized between sites, and workloads can be redistributed among them. If one site goes down, another automatically takes over the traffic. To the end user, this appears as uninterrupted service operation.

This architecture allows addressing two different classes of tasks that are often confused.

High Availability — protection against unforeseen incidents. If one site is unavailable, the service continues to operate on another, minimizing downtime.

Disaster Recovery — protection against more severe scenarios: data loss, prolonged site unavailability, or physical destruction of equipment. DR (Disaster Recovery) enables recovery after a major incident.

How It Works Technically

The foundation of geographically distributed architecture is the Data Center Interconnect (DCI) technology. Its essence lies in creating secure, high-speed links between sites, through which data is transmitted, virtual machine states are synchronized, and traffic is routed.

DCI links are laid out along independent physical routes. If one becomes unavailable, traffic is automatically switched to a backup. The switchover occurs without session interruption and without packet loss.

Data is synchronized between sites in real-time or with minimal latency. This means that when switching to a backup site, the business does not lose transactions or revert to a state from weeks ago.

The effectiveness of such an architecture is measured by RPO (Recovery Point Objective) — the maximum acceptable data loss. This metric directly depends on the frequency of synchronization. The better the architecture is designed, the lower the RPO; ideally, it approaches zero, indicating minimal data loss. 

Meanwhile, recovery time with proper configuration takes just minutes. 

In April 2025, a new point of presence was added to our network in Almaty — Sairam Data Center, expanding possibilities for geographically distributed configurations in the region.

Who Actually Needs Geographically Distributed Infrastructure

In reality, geographically distributed infrastructure is needed by almost everyone. Large businesses critically require such an infrastructure architecture, and in the Kazakhstani market, most providers have already built their infrastructure according to a geo-distributed scheme. However, if we were to divide them, then: 

— Geo-distribution is necessary for companies in the financial sector, retailers, e-commerce businesses, healthcare companies, firms with customers nationwide, and businesses with strict SLAs.

— Geo-distribution makes sense when the cost of downtime exceeds the cost of redundancy. For some businesses, the balance favors a single site.

For instance, a startup at the MVP stage may manage without geographically distributed infrastructure. The same applies to test and development environments — in these cases, geo-distribution is usually not critically necessary.

This can be assessed simply: if an hour of downtime costs you more than a month of infrastructure redundancy, you should choose redundancy.

How to Choose a Provider

The effectiveness of a disaster recovery solution is determined not only by the geographic distribution of sites. When auditing a cloud provider, we recommend checking the following aspects: 

1. The provider has availability zones in different cities and countries. If the provider’s sites are powered by the same substation, there is little point in this from a high availability perspective.
2. TIER-III data centers. This level guarantees the ability to perform technical maintenance and repairs without stopping IT equipment.
3. DCI link bandwidth. The higher the load and the stricter the RPO requirements, the wider the channel between sites should be. Otherwise, it becomes a bottleneck during peak loads or mass data replication.
4. Support for required replication technologies. It is important to ensure that the provider supports technologies compatible with your stack. For example, for VMware infrastructure, these are vCloud Availability and vSphere Replication.
5. Transparent SLA with financial accountability. Availability guarantees must be documented in the contract with specific metrics and consequences for non-compliance.

A properly designed architecture with multiple points of presence, DCI links, and automatic failover allows transferring infrastructure risks into a manageable category. The business continues to operate regardless of what happens at a specific site.