Why do network claims matter?
In SASE, traffic travels from a user or site to the nearest point of presence (PoP), gets inspected, and then travels on to its destination. The distance to that PoP, the path between PoPs, and the capacity at each location all shape what users experience. That is why the network carries 15% of our total.
What do the vendors state?
Cato states a "cloud-native, private global backbone spanning 100+ PoPs worldwide", branded Cato Neural Edge, with traffic carried between PoPs on that backbone. Fortinet states "a global network of 200+ PoPs" on Fortinet-owned infrastructure, extended through AWS, Google Cloud and Oracle Cloud. Netskope states its NewEdge network runs on 120+ data centers in 80+ regions, with all SASE services in every location. Palo Alto Networks, Zscaler and Cisco do not give a count on the pages we read.
Source: Cato global private backbone
Source: Fortinet FortiSASE
Source: Netskope NewEdge
Why can't you compare the counts?
A PoP in one vendor's language may be a data center in another's, and a region may hold several data centers. Some counts include locations hosted on public cloud, some do not. A larger number does not tell you whether a location near your users runs every service, or only some. We quote each figure in the vendor's own words and do not rank them.
What should you ask instead?
Ask for the list of locations nearest your users and sites, and which services run at each. Ask who owns and operates the hardware. Ask how traffic moves between PoPs: over a private backbone the vendor runs, over the public internet, or over a cloud provider's network. Ask what the uptime SLA covers and how credits work, and get the SLA document itself, since uptime figures on marketing pages rarely carry the terms.
What about uptime figures?
Cato and Palo Alto Networks both state a 99.999% uptime SLA on the pages we read. Neither page includes the SLA terms. Treat the figure as the start of a contract conversation.
Source: Cato global private backbone
Source: Palo Alto Networks Prisma Access
How should you test it?
During a proof of concept, measure latency from your main sites to the vendor's nearest location and to your key applications, at busy hours. Compare the path with and without the service. Your own numbers are the only network figures that are comparable across vendors.
How does the network affect overhead?
A vendor that runs its own backbone takes on routing between locations, so your team manages less. A vendor that relies on the public internet between locations may ask you to keep more of your own wide-area network. Neither is wrong, but the split of work should be clear before you sign. Ask which parts of the path the vendor monitors and which parts you still own.