Bigger logs.
100×
-
The size of logs has increased by 100x.
Agent traces are larger than typical server logs.
-
Probabilistic services require more logging.
You can’t predict them, so you must observe them.
-
Evals don’t cover it.
You cannot anticipate how users will interact with your service. The model behind the API could be changing. You need to continuously monitor reality.
-
Mistakes and errors must be explained.
Your customers will require it. Your engineering team will require it.
Reliable performance.
10.6 ms
Logmoth average
491.8 ms
Logmoth average
Common anxieties.
No multi-tenancy.
Each instance runs on dedicated compute. You share nothing with other customers.
Yes, it's more expensive, but by O($50/month). Compute is not the bottleneck.
How billing works
Choose a plan by CPU and memory. Compute is prorated to the second while your instance runs; storage is billed separately. Each month, you’re charged for the previous month’s usage.
The calculator above estimates storage only. Your plan determines the compute price.
- US data jurisdiction Yes, ofc
- Logs are stored and processed in the United States. Choose US East (N. Virginia) or US West (N. California).
- Separate read and ingest keys Yes, ofc
- Give producers ingest keys and readers read keys. Ingest keys cannot read log events; read keys cannot write them.
- Zero-downtime key rotation Yes, ofc
- Keep multiple keys active during a rollout. Create a replacement, move clients to it, then revoke the old key without interrupting ingestion or queries.
- SOC 2 + signed DPA Not yet
- SOC 2 and a signed data processing agreement (DPA) are not yet available.
Principles
-
Built on commodity, infinite-scale object storage
-
Stable query performance at any scale
-
Separates large log payloads from queryable attributes
-
Low memory and compute requirements
-
Simple, opinionated structure to prevent sharp edges