Overview
A SaaS provider's telemetry ingest was growing well ahead of its business, with no clear owner for the growth. AceMQ helped them attribute volume, cut what nothing consumed, and put controls at the collection layer so the reduction holds.
Challenge
Ingest was dominated by a small number of sources: verbose framework logs forwarded at debug level, custom events emitted per request that were only ever aggregated, and infrastructure metrics collected at a much finer interval than any dashboard displayed. Nobody could see which team's telemetry drove which share of the bill, so cost conversations went nowhere and the growth continued each quarter.
Environment
New Relic across a multi-account AWS estate with Kubernetes workloads and several autonomous product teams.
Approach
AceMQ does not operate the vendor platform — we help customers decide and control what leaves their infrastructure. The engagement attributes ingest by team and source, checks what is genuinely queried, and implements drop rules and sampling at the collection layer where they apply regardless of what individual applications emit.
Solution
- Ingest volume attributed by team, service, and telemetry type so cost has an owner rather than a total
- Query and dashboard usage cross-referenced to identify telemetry that is ingested but never consumed
- Drop rules applied at the collection layer for provably unconsumed log and event classes
- Log forwarding levels corrected at source, removing debug output left enabled from past incidents
- Infrastructure metric collection intervals aligned to the resolution dashboards and alerts actually use
- Per-team ingest budgets with monitoring so a logging change surfaces as an alert rather than an invoice
Outcome
Ingested volume fell by roughly forty percent with no dashboard or alert losing its data source. Because ingest is now attributed per team and budgeted, growth is a visible decision rather than a quarterly surprise.
Technologies
New Relic
OpenTelemetry
Kubernetes
Terraform
Java
Linux
