← All posts

SASE Architecture for International Sites

Night-time aerial view of an illuminated coastal region with connected towns, cover image for the article SASE Architecture for International Sites

A company with sites in several countries asks different questions of SASE than a business with a single office building. Every site and every mobile team should get the same protection, without applications slowing down because traffic runs through headquarters. On top of that come rules about where data may be processed, and countries with network conditions of their own.

This article collects the criteria that decide the choice of a SASE architecture for international sites, along with the questions you should ask vendors about them. What SASE/SSE is in principle is explained in our article What is SASE/SSE?. How the rollout works is covered in our five-phase guide.

What SASE has to deliver across borders

In its publication SP 800-215, NIST describes SASE as an architecture that converges networking and security functions and delivers them as a cloud service at global scale. Gartner coined the term in 2019. The services are reached through distributed points of presence (PoPs). Sites, data centers, cloud environments, and end devices connect to a PoP, and that is where policies are enforced. Our glossary explains the terms SASE and SSE.

For a single site, the discussion is mostly about features. With many countries, three questions come on top: How close is the nearest PoP to every site and every mobile team? Where are traffic and logs processed? And who runs the whole thing across time zones?

PoP coverage and latency at distant sites

For any site, a SASE service is only as fast as the path to the nearest PoP that runs all the services it needs. So don't just ask how many PoPs a vendor has. Ask which services are available where. In its SASE reference architecture, Cloudflare states that all Cloudflare One services are designed to run in every location of its network. Traffic is routed via anycast to the closest data center, so inspection and policy enforcement happen close to the source.

Vendor claims are no substitute for measurement. During the proof of concept, test latency from your actual sites and home offices, especially in regions with few users. With Cloudflare One, Digital Experience Monitoring (DEX) helps here: synthetic tests run periodically from end devices and show whether a problem lies with the device, the network, or Cloudflare.

Some SaaS providers and partners only accept connections from allowlisted IP addresses. At Cloudflare, dedicated egress IPs are tied to a single data center. If your users are spread across several regions, Cloudflare recommends multiple egress IPs in different data centers, each assigned to the closest user group.

The special case of China

For users in mainland China, Cloudflare notes that traffic routed through servers outside the country faces significant latency and reliability issues due to China's network boundaries. The China Network therefore runs selected Cloudflare products in data centers in mainland China, operated by its partner JD Cloud. For SASE, the Global Acceleration add-on matters: it enables the Cloudflare One Client and Cloudflare WAN inside China and is provided by partners such as China Mobile International, CBC Tech, and JD Cloud. A travel SIM is available for employees visiting China.

Clarify early which contracting partner covers your China sites and which features are available there. According to Cloudflare's documentation, for example, Gateway does not support dedicated egress IPs on the China Network. Cloudflare also points out that content in mainland China is monitored by local authorities and must comply with local regulations.

Data residency: look at processing and logs separately

When you ask where data is processed, SASE rarely gives a single answer. For Cloudflare One, the documentation on data residency explicitly separates three layers:

  • Connection: According to Cloudflare, the tunnel from a client or site may terminate in a data center outside the selected region. The HTTPS traffic inside stays encrypted until Gateway forwards it to the selected region for inspection.
  • Processing: With Regional Services for Gateway, you define the region where Gateway decrypts and inspects HTTPS traffic, for example the European Union or the Schengen Area.
  • Logs: The Customer Metadata Boundary keeps supported logs stored in the EU or the US.

The details matter. Gateway DNS and network policies are processed in the data center where the traffic arrives; the processing region does not apply to them. CASB supports neither regional processing nor regional log storage, File Sandboxing processes files in the US, and Browser Isolation supports regional processing only in the US FedRAMP region. With the Customer Metadata Boundary turned on, Gateway logs do not appear in the dashboard and are exported through Logpush. The Data Localization Suite is a paid add-on for Enterprise customers.

Ask every vendor for this kind of breakdown per feature instead of a blanket promise. The legal assessment of your processing stays with your data protection officer, and the breakdown gives them the basis for it.

Single-vendor or multi-vendor

In a single-vendor approach, site connectivity and security services come from one provider, with one console and one rule set. In a multi-vendor approach, you might combine an existing SD-WAN from one manufacturer with Security Service Edge (SSE) from another and connect the two through IPsec or GRE tunnels. According to Cloudflare's reference architecture, most organizations move to SASE progressively, and some deliberately use several vendors. For most, it says, the aim is to consolidate security with a single vendor to simplify management and visibility.

Both paths are legitimate. One vendor means fewer interfaces and no duplicate policy maintenance. Several vendors let you keep an SD-WAN that works, but they bring two consoles, two support channels, and an interface someone has to own. So the label matters less than who troubleshoots when something breaks between the site and the cloud.

SD-WAN and site connectivity

Sites reach a SASE platform in different ways. At Cloudflare, these include the Cloudflare One Appliance, which sets up IPsec tunnels on its own, as well as IPsec or GRE tunnels from existing routers, firewalls, and SD-WAN devices. The tunnel endpoints are announced via anycast, so traffic from a site arrives at the closest active data center. If that data center is unavailable, Cloudflare says traffic is automatically rerouted to the next-closest one.

If you already run an SD-WAN, check whether a validated guide exists for your devices and which limitations it lists. One example is the Cisco Meraki MX integration with Cloudflare WAN. It requires at least firmware MX 19.2.7, which older hardware may not be able to run. The redundant tunnels operate in active/standby mode. The guide distinguishes three variants: only traffic between sites, only internet traffic through Gateway, or both through Cloudflare.

The SD-WAN remains responsible for the circuits: which connections a site has, what redundancy looks like, and which provider delivers in which country. What matters there is summarized in our checklist for migrating from MPLS to SD-WAN.

Identity and device posture

Across borders, one rule holds: a device's location says nothing about whether it can be trusted. NIST SP 800-207 sets out as a Zero Trust tenet that access to resources is decided by dynamic policy that includes, among other things, the observable state of the client identity and the requesting asset. For your selection, this means the platform has to connect your identity providers and factor device posture into every access decision, for example when users reach applications through Zero Trust Network Access.

If subsidiaries run their own directories, it matters whether several identity providers can be connected in parallel. Cloudflare One supports SAML- and OIDC-compliant providers, including Microsoft Entra ID, Okta, and Google Workspace, and several at the same time. The client checks device posture for things like disk encryption, operating system version, or a running firewall. Solutions such as Microsoft Endpoint Manager, CrowdStrike, or SentinelOne can be integrated as well. According to Cloudflare, a change in a device's posture affects its access immediately.

Running it in-house or as a managed service

A SASE platform for many countries is not a project with an end date. Policies change, applications are added, sites open or close, and users call in at times when nobody is working at headquarters. So before you choose, settle who handles which tasks: policy maintenance, exceptions, monitoring, updates, provider management, and ticket handling.

Running it in-house, your team needs know-how on the platform and on Zero Trust, plus on-call coverage that matches the time zones of your sites. A managed service shifts this work, but it requires clear responsibilities, for example in a RACI matrix. Also ask who is accountable when a fault sits between the SD-WAN and the SASE platform.

Migration path: country by country instead of a cutover date

With several countries, switching on a fixed cutover date is an unnecessary risk. In our projects, this sequence has proven itself: first a pilot with real users, then parallel operation with log comparison, then the switch country by country. The old environment stays in place as a fallback until the new one has proven itself in daily use. That is how we approached a project with sites in twelve European countries. The full process is described in our guide to adopting SASE in five phases.

The criteria at a glance

The table summarizes the criteria and gives one question for vendors on each.

Criterion

What to look for

Question for the vendor

PoP coverage and latency

Proximity to all sites, all services at every PoP, your own measurements

Which services run at the PoPs near our sites, and can we measure there?

China and special regions

Local partners, separate contracts, limited features

Which features are available in China, and through which contracting partner?

Data processing

Region for decryption and inspection, per feature

In which region is which traffic decrypted and inspected?

Logs

Storage location, access, and export

Where are logs stored, and how do we export them?

Platform model

One vendor or a combination of several

Who owns the interface between SD-WAN and security services?

Site connectivity

Validated guides for existing devices, redundancy

Is there a validated guide for our devices, and which limitations does it list?

Identity and devices

Multiple identity providers, device posture, endpoint solutions

Can all of our identity providers and endpoint solutions be connected?

Source IPs

Egress IPs per region for allowlists

How many dedicated egress IPs do we need, and in which regions?

Operations

Service hours across time zones, clear responsibilities

Which tasks stay with us, and which response times apply?

Frequently asked questions

What criteria matter when choosing SASE for international sites?

What matters is how close the PoPs are to all sites, the region for decryption and logs, how the existing SD-WAN connects, identity providers and device posture, and operations that work across time zones. Countries such as China add network and contract conditions of their own.

How does SASE affect performance at distant sites?

It depends mainly on the path to the nearest PoP that runs all the services you need, because traffic no longer has to go through headquarters. At Cloudflare, clients and site tunnels are routed via anycast to the closest data center. A measurement during the proof of concept shows whether that is enough.

Do we have to replace our existing SD-WAN for SASE?

Not necessarily. Cloudflare WAN accepts IPsec or GRE tunnels from existing routers, firewalls, and SD-WAN devices and documents the integration for various manufacturers, including Cisco Meraki MX. Check the prerequisites and limitations in the relevant guide before you settle on a target design.

Where does a SASE platform process data?

It depends on the feature. With Cloudflare One, you set separately the region where Gateway decrypts and inspects HTTPS traffic and where supported logs are stored. Some features, such as CASB, do not support regional processing. So get the situation in writing for each feature.

How KAEMI helps

We implement SASE/SSE with Cloudflare One and connect sites with SD-WAN based on Cisco Meraki. As a Cloudflare partner, we support you from architecture consulting through design, implementation, and migration to the managed service. In an analysis workshop, we look at your sites, applications, and access paths and show what a SASE/SSE model for them can look like. In operations, we handle tuning, updates, exceptions, and reporting for Cloudflare One, and for SD-WAN also monitoring, fault resolution, and provider management. We configure Regional Services and the Customer Metadata Boundary to your requirements and document the settings for your compliance records. Read more on our Secure Access Service Edge page, or contact us directly.

Sources: NIST SP 800-215, Guide to a Secure Enterprise Network Landscape; NIST SP 800-207, Zero Trust Architecture; Cloudflare: Evolving to a SASE architecture; Cloudflare: Data residency for Cloudflare One; Cloudflare: Data Localization Suite; Cloudflare: China Network; Cloudflare: Global Acceleration; Cloudflare: Dedicated egress IPs; Cloudflare: Cisco Meraki MX with Cloudflare WAN; Cloudflare: Digital Experience Monitoring.

Want to secure access consistently with Zero Trust?

Related to “SASE Architecture for International Sites”: KAEMI designs, implements and manages SASE/SSE with Cloudflare One: ZTNA instead of VPN, verified access from anywhere — as a managed service.