Blog post
Recently, Catchpoint expanded our Global Agent Network to over 3,000 agents. In a crowded space, this is by far one of our key differentiators. At the time of writing, no one else boasts 395 providers in 105 countries and 346 cities. As Director of ISP Strategy, I’m not here to pat myself on the back—my real question is: why? Why build such a massive, independent network — going through all the effort to place backbone agents in hard-to-reach regions like China, Russia, and several African countries?
The answer lies in how the Internet is built. It isn’t a single, monolithic network but a patchwork quilt of thousands of independent networks—ISPs, data-centers, backbones, wireless carriers, and more—all stitched together by peering and transit agreements worldwide. To monitor performance accurately, you need visibility into every layer of that quilt, or, what we like to call the Internet Stack.
In this article, we’ll unpack the five types of Catchpoint synthetic agents—backbone, wireless, last-mile, cloud, and BGP—and show you exactly when and why each matters for keeping that quilt intact. First, let’s explore how the Internet actually connects end to end.
Why multiple vantage points are key
The “Internet” isn’t a single cloud—you can’t just tap into one place and see it all. It’s really tens of thousands of independent networks (ISPs, data centers, wireless carriers) stitched together by peering and transit deals. In a peering arrangement, two networks exchange traffic for free. In a transit relationship, one network pays another to carry its traffic. Peering keeps traffic local; transit carries it farther afield.
Because each network makes its own choices about peering and transit, you need monitoring agents at many points to see what’s happening. A performance hiccup in one ISP’s peering location might not show up at a different ISP’s vantage point. That’s why Catchpoint places agents in dozens of key networks—so you won’t miss an issue that affects only a slice of the Internet.
A map of how thousands of networks (Autonomous Systems) peer and buy transit around the world. Source
Why tiers matter
All those peering and transit agreements naturally sort networks into tiers:
- Tier 1: These are very large networks that peer with each other and don’t need to buy any transit at all to reach any corner of the Internet. They are considered the backbone of the Internet as they will typically carry long-distance Internet traffic. While the networks belonging to this group have changed since the beginning of the Internet, it’s remained relatively stable, including providers such as Lumen, AT&T, Cogent, Verizon, Orange, GTT, NTT, and Telxius. This category features very large traditional telecom providers that have been serving their domestic market for many years.
- Tier 2: Regional providers that both peer and buy transit. The scale of operations as well as the type of services they provide (IP transit, Ethernet or Dark Fiber wavelengths) will dictate the number of peering connections and transit providers. Most networks will fall into this category if they peer at one or more Internet exchange points and have two or more upstream providers.
- Tier 3: Smaller, local ISPs (often single-homed) that feed to an upstream provider. They show you what your end users see on a residential or localized network.
Putting agents in each tier matters because a Tier 1 network will have more visibility into global events (total or partial outages, backbone congestions, etc) as opposed to a regional Tier 2 or a local Tier 3 network. On the other hand, localized outages affecting a limited number of providers in a particular geographical area, won’t be easily observed unless having visibility from one of the affected networks.
Why single-homed Tier 1/Tier 2 connectivity matters
Now that we know how networks sort into tiers, let’s look at how those tiers influence the way we connect our agents.
Many data centers, hosting and managed service providers typically use multiple upstream ISPs (Tier 1 and Tier 2) to create a single, aggregated connection to the Internet. This Internet connectivity, normally offered as a service, is often called blended bandwidth or multihoming.
Multihoming improves redundancy as traffic can be rerouted if one ISP goes down or has packet loss/congestion. It also improves performance as different ISPs may offer better latency to different geographies.
But for Internet Performance Monitoring (IPM) however, the use of multihomed agents instead of single-homed Tier1/Tier2 carriers introduces variability in your monitoring data, making it harder to identify and troubleshoot the issues affecting performance.
Here is why more than 96% of Catchpoint backbone agents use single-homed Tier 1/ Tier 2 connectivity instead of blended bandwidth:
- Path consistency: A consistent Tier 1 upstream path reduces variability, making anomalies and degradations easier to detect and attribute.
- Backbone visibility: Tier1 visibility is essential to observe how the core Internet behaves in relation to routing anomalies, BGP hijacks or backbone congestion.
- Performance stability: With blended connectivity, routes may change dynamically based on load-balancing or pricing strategies (e.g., BGP-based traffic engineering), affecting performance results.
Now that you understand how and why Catchpoint chooses single-homed Tier 1/Tier 2 connectivity, let’s look at each of our five synthetic agent types. We’ll explain where we place them, how we build them, and—most importantly—exactly what visibility each one gives you.
Backbone agents
Backbone agents give you a “core-of-Internet” vantage point to catch global outages, BGP hijacks, and CDN-level issues no other agent can see.
We place backbone agents in Tier 1 or Tier 2 ISPs worldwide, selecting carriers by:
- Geography and market importance (global connectivity hubs)
- CAIDA ASRank & APNIC eyeball data (to cover the most interconnected networks)
Each backbone agent runs as a server cluster in a carrier-neutral data center with dedicated IP transit. Carrier neutrality ensures multiple international and domestic carriers via cross-connects, while colocating servers in one facility reduces colocation costs. In emerging markets (e.g., parts of Africa or China), where neutral data centers are scarce, we may host clusters in carrier-owned facilities—only as a last resort.
Measuring performance and availability from backbone agents is critical for:
- Experience Level Objective (XLO) measurements: Validate service performance when source and target share the same ISP.
- CDN performance/validation: Ensure fast, reliable content delivery across the backbone.
- Competitive benchmarking: Compare your service to peers in the same Tier 1/2 networks.
- Peering/ISP monitoring: Detect routing changes, BGP anomalies, or unexpected transit behavior.
- Geo-based DNS validation: Confirm DNS resolution speed and correctness from the core network
Public cloud agents
Cloud agents give you visibility right inside public-cloud data centers—so you can catch platform-specific issues before they impact users.
We run 280+ cloud agents across every key availability region in AWS, Azure, Google, Oracle, Alibaba, Tencent, Akamai Compute, and OVH.
Measuring performance and availability to and from cloud agents is essential if you are hosting applications in the cloud or using any of their computing products. Cloud agents allow your SRE teams to pre-emptively detect performance degradations on public clouds that can affect how your users are experiencing your applications and services.
Wireless agents
Wireless agents simulate real-world cellular conditions, giving you a true picture of how your applications perform on 3G/4G/5G networks.
We place wireless agents using AWS Wavelength and independent carriers in the US, Canada, Japan, Germany, Korea, India, and the UK (e.g., Verizon, KDDI, BT, T-Mobile 5G, AT&T 5G).
Running wireless tests alongside backbone tests lets you compare mobile experience to core-network performance—so you can spot issues like packet loss or DNS slowdowns that only affect cellular users.
Last-mile agents
Last-mile agents live in real homes, giving you a true end-user view of broadband performance.
Our last-mile agents run on small customer-premise devices that connect to a residential ISP. Use these agents to troubleshoot ISP-specific issues—like throttling, DNS failures, or regional outages—that only affect subscribers on a particular network.
Enterprise agents
Enterprise agents give you visibility into your own network—from branch offices to data centers to edge locations.
Enterprise agents are deployed within your organization’s infrastructure. That includes office networks, private data centers, retail locations, or edge devices. These agents help you monitor internal applications, APIs, and services with the same level of granularity you get for external traffic. Combined with our Global Agent Network, enterprise agents complete the picture—giving you visibility from both outside-in and inside-out.
BGP agents
BGP agents watch the real-time routing table, so you can catch hijacks, leaks, or unexpected path changes that threaten your service.
Catchpoint maintains a route collector infrastructure that process real-time routing data from 1700+ BGP agents with the goal of monitoring BGP activity and detecting issues such as route hijacks and leaks.
In addition to using RIPE RIS and RouteViews datasets, Catchpoint operates its own private collector infrastructure which includes agreements to receive data from 330+ BGP agents from 100 unique networks.
If you share your own BGP sessions with our private collectors, you’ll gain even deeper insights in your portal—so you see exactly how routing anomalies affect your prefixes.
Wrapping it up
When I asked, “Why build a network of over 3,000 agents in 105 countries and 346 cities?” The simple answer is that today’s Internet isn’t one giant cloud but a patchwork quilt of independent networks.
By spreading our agents across every layer of the Internet Stack, we can reveal problems at the very moment they start—whether it’s a routing change in a distant backbone, a subtle slowdown in a public cloud region, or a local ISP hiccup affecting a handful of homes.
This broad visibility isn’t about boasting coverage; it’s about ensuring that whenever something goes wrong, you know exactly where to look. In other words, the effort we put into building and maintaining such a diverse network isn’t just a technical feat. It’s the key to preventing those 3 a.m. wake-up calls for your IT team, avoiding frantically assembled war rooms, and keeping your users happy no matter where they connect.
Ready to unlock the full potentialof observability?
Learn more about Catchpoint'sintelligent agent network and how it can transform your monitoring strategy: https://www.catchpoint.com/global-observability-network
This is some text inside of a div block.
