How the ProbeNet PoP acquisition works (or why there is no ceiling on how many nodes we want).
ProbeNet is our proprietary network of distributed compute nodes across 550 cities in 150 countries. It is an ever-growing massive network dedicated to running internet measurement to generate first-party raw network measurements. Currently, it stands at about 1320 nodes (January 22, 2026)
How we decide to procure PoPs
The ProbeNet network favors two things:
- Network diversity
- Location diversity
Actually, network diversity includes location diversity.
Network Diversity
The idea of network diversity involves low-latency data we have for a particular ASN. If we host a node in a particular ASN, naturally the latency within that ASN will decrease, as will the latency with nearby peering ASNs.
However, if the ASN operates across multiple regions, hosting a PoP in one country will not result in low latency to the ASN’s IP addresses in another country.
So, although we would like to host a node in every single ASN, we are actually targeting ASN multiplied by Location.
We aim to reduce latency metrics for entire ASNs, regardless of the number of PoPs we need to host across the locations they operate.
This information is monitored internally, and we have targets to reduce global latency for “every” IP address. Our goal is to ensure that X% of all ASNs are within X ms from a ProbeNet PoP. The lower the latency data across all IP addresses leading to less noise, the better our data performs.
Internally, we flag the ASN, the ranges, and other metadata. Our Founder & Co-CEO primarily oversee foundational data queries and discovery efforts (from data point of view), after which we reference the data and act on it.
Simultaneously, we build internal apps to monitor performance. However, we continuously identify new network efficiency opportunities and have many internal dashboards for investigating and procurement. This discovery process is a collaboration between the data and the research team.
Location Diversity
Although the “X” country and “X” cities sound quite interesting, it is a very small piece of the puzzle.
Our primary goal is to reduce ProbeNet latency globally. It is a continuous effort. So, having nodes across different regions improves measurement data. The primary goal is to host a single PoP in each country, on every major cities and on every island. Because hosting a probe within a geographical region will reduce the latency measurements we have for that region.
Currently, we are aggressively investing in discovery and deal-making across several dozen countries, many of which are Pacific Island nations, smaller EU regions (less than 50k in population), and central African countries.
As we rent and self-operate these servers, we have to navigate through complex international procurement and business operations processes, which is just standard operating procedure for us.
Procurement Process
ASN latency
If we have low latency across all the IP address range for a particular ASN, we will not procure the server. This could be because we already have a server with the ASN across all the locations they operate or we have several servers that peer with the ASN.
Considering we have ~1320 (January 22, 2026) servers across all major data centers, we have extremely low latency to publicly known ASNs that provides commercial hosting services.
However, the goal is “overall latency”. So, that means we also want to bring down the ASN latency of ISP, business, educational insititute or even government ASNs. So, it is a continous process to help us bring our ProbeNet latency measurement down.
Another case is that if a paricular ASN operates in many different locations, even if we have an node with them we will look into hosting more nodes to bring the whole ASN level latency down.
IPv6
We genuinely believe in IPv6 and we want that to be reflected in our data. We actively look for hosting provider partnerships that support IPv6 connections to our node.
Geographic region diversity
We monitor country level data as well. We look into country data and cities with high population, significance, and act as network hub. We focus on bringing at least one server there.
Even though there could be a backhaul networking issue which results in high latency even if you are pinging a server next to you, we will try our best to ensure we collect the data and register the behavior of the network topology. ProbeNet is not about singular network measurements, it is about mapping the internet topology.
Collaboration with the community
This is not our primary method of operation, but many of our users and community members themselves operate ASNs, ISPs, and Data Centers. We have been providing accurate Internet data for nearly a decade and a half now. They have the hardware and capacity to support us, and because they like us, they often offer us their hardware through data sponsorship. We are grateful for their generosity.
Our community of users themselves volunteer to help. Our community encompasses teenagers running their hardware for Minecraft servers all the way up to enterprise tech leadership and government agencies. We always actively try to support their data needs through open data access, so sometimes, through the generosity of their hearts, they would offer us sponsored servers. We only have a handful of sponsored servers, but we appreciate our community’s generosity in supporting us.
Partnerships
With academia and major internet organizations, we are spearheading high-level cross-collaboration partnerships led by our partnership team. We aim to focus on mutually beneficial data center collaboration on a global large scale for enterprises and research/academic institutes.
What is our ultimate target number
The ProbeNet does not have a fixed number of PoPs. It is about the latency to all IP addresses.
We have no idea how many probes overall global latency to IP addresses in our data will go down. We are not talking about an average global latency number here. The idea is about an all-encompassing, comprehensive global latency number. If there are edge cases of high latency, we are not going to ignore them; we need to bring those numbers down.
At what number of ProbeNet PoPs will we be satisfied? We have no clue. Maybe 5,000 or maybe 10,000. There are 70,000 ASNs out there, hundreds of thousands of cities, and billions and billions of IP addresses. Our data is to remove measurement noise in latency.
It is an ever-evolving, never-stopping effort to scale. The procurement process is getting harder, and we think we have reached the point that we have tapped into every self-serve hosting service. Now, procurement processes involve national and local ISPs, and a lot of door-knocking at ASNs. But still, as this is the cost of good data, we are happy to continue investing in it.