People read dashboards because the logs were too large to trace. Agents follow the log and act. FFWD keeps that memory searchable, with anomaly intelligence, inside your environment.
Keep recent data you need investigating close at hand. Extend your telemetry history into open tables on object storage you own and control.
Short-term · Parquet · Arrow · in memory
A Parquet store with in-memory Apache Arrow search. Telemetry is immutable and time-indexed, so ingestion stays fast and schema is applied when you query.
Apache Iceberg + your object storage
Extend retention into open Iceberg tables, with catalog integration to make historical telemetry discoverable and queryable.
People left logs for dashboards because the volume was too much to trace. Agents are good at exactly that trace, and a metric value does not give them an action. Logs land here as Parquet and are searched in memory with Apache Arrow.
Logs are the common format. Structured or raw, if they are time-indexed they ingest fast. Quiver is tuned for that path, not for a general database.
Typical time-series stores slow down as metric cardinality grows. Quiver does not. Log search stays in memory, in Apache Arrow, and anomaly detection runs on those logs.
Explore anomaly intelligence →Agents reach the logs through MCP. PromQL and SQL stay open for third parties, such as a Grafana dashboard or another system, to query the same store directly.
Long-term telemetry is written as Iceberg tables on object storage you own. FFWD hosts the Iceberg catalog in Rust, the same stack as Quiver, so those tables are open to an external query engine as soon as they are stored.
The Iceberg connector writes tables to object storage. Use more than one destination, and more than one table in each bucket, so retention matches how you want the history kept.
Iceberg catalog runs inside FFWD, in Rust. It serves the catalog for external queries, so the tables on object storage are discoverable rather than files sitting in a bucket.
Logs dumped as files have to be pulled back out and transformed before anyone can analyse them. Iceberg tables are already tables. Bring a query engine and the history is ready.
Agents cannot work through terabytes of raw logs. FFWD filters that down to a signal, keeps the recent evidence fast in Quiver, and leaves the older history queryable in Iceberg. Agents reach both through MCP.
Search the logs already in fast storage, with anomaly evidence beside them. The useful signal, not the raw lake.
Older logs are Iceberg tables, not files waiting to be rehydrated. Bring a query engine and the history is ready.
The agent traces the logs through MCP and gets to the cause sooner. Time to recovery gets shorter.
Explore agent assurance →