Segment 4
In this sub-session, we first investigated apparent data corruption detected during S3 load testing, adding better error classification to the loadtest tool and confirming that the 'verify errors' were actually context deadline timeouts rather than real corruption. We then focused on optimizing the YCQL write path to improve throughput by implementing a `CQLBatcher` in the `database/cqldb` package that collects individual CQL INSERT calls and flushes them in batches (default 15,000 entries or within 10–30 ms) using a worker pool with exponential backoff retries, integrating it into `ObjectIndexCql.Put` via a new `Session()` method on the `Database` interface. Additionally, we fixed a configuration bug (`RetrievableRepairThreshold > MinimumReplicaCount`) that prevented kuri nodes from starting and updated the `gen-config.sh` script. To eliminate the Docker userland proxy as a bottleneck at high concurrency, we rewrote `docker-compose.yml` to use host networking and removed the unnecessary nginx web‑UI proxy. Load tests with the batcher showed clean results at 10 workers (~115 MB/s, 0 corruption) and throughput scaling to ~334 MB/s at 100 workers, though connection resets appeared due to the proxy; the host network change is expected to resolve these issues and allow testing at much higher concurrency.