Top Hosts
See which external hostnames your app talks to most, ranked by request volume with p95 latency and error counts to spot misbehaving third-party APIs.
Top Hosts shows which external hostnames your app talks to most frequently — your own APIs, third-party services such as payment gateways and SMS providers, CDNs, analytics endpoints. Each host is ranked by request volume and shows p95 latency + error count so you can instantly spot a misbehaving third-party that's silently dragging your UX down.
Only outbound HTTP requests appear here. Database queries, custom operations, and cache spans don't appear here.
What You See#
For each host:
- Host — e.g.
api.payments.example.com,cdn.example.com - Requests — total for the selected time range (hosts are ranked by this)
- Average and p95 latency, in milliseconds
- Errors — requests that returned HTTP 400 or higher
To drill into slow or failing outbound calls, use the Operations and Failed requests cards on Trace → Performance in Abrova Console.
What Counts as a Host#
The host is automatically extracted from the request URL by every official SDK. For example, a request to https://api.payments.example.com/v1/charges is bucketed under api.payments.example.com. Path differences don't matter at this level — that's what URL Performance is for.
Common Patterns to Watch#
Healthy if it's your own API. Suspicious if it's a third-party — could mean retry storm or over-eager polling
A 1.5s p95 from a payments or auth host means your checkout / login is gated on it. Cache or async if possible
Usually a new feature shipped that calls an unfamiliar service. Verify it's expected — could be a misconfigured SDK
A small error count from your CDN often signals regional outages users will report before you notice from logs
Real-World Use Case#
A common scenario: your app's overall p95 spikes from 300ms to 1.1s overnight. The latency chart confirms it, but the Operations card looks normal — every operation is similarly slow. Top Hosts immediately reveals the culprit: api.thirdparty.com p95 jumped from 200ms to 2.4s. Their incident is now your incident — but at least you know it's not your code, and you can mitigate (cache, async, fallback) rather than chase a ghost.
Best Practices#
Track their p95 like you track your own. A slow payment gateway or SMS provider is your problem the moment users notice
Top Hosts shows aggregate health; Failed Requests shows specific failures. Use both together
For hosts with consistent high TTFB but low p95, a <link rel="preconnect"> can save 100-200ms on the first call
A brand-new host appearing near the top of the list is usually either a planned launch or an exfiltration risk — verify either way