HANA write performance degrades every few minutes. Traces show savepoint duration climbing and users see short write stalls, while CPU and network look normal. Which layer should be investigated first?
Show answer and explanation
Answer: B
During a savepoint HANA writes modified pages from memory to the data volume and persists savepoint information, so storage that cannot absorb the burst stretches the savepoint and makes user transactions wait. The data volume metrics to review are write operations and bytes, queue length, and whether the volume is hitting its baseline throughput or exhausting burst balance, together with the file system and LVM layout. CPU and network can look healthy while storage is the real bottleneck.
Why the other options are wrong
- A. Inter-zone latency affects replication to the secondary and log shipping, not the local savepoint write to the data volume.
- C. DNS resolution happens when a connection is established; it would not produce a periodic stall aligned with savepoint activity.
- D. A CDN cache ratio affects static content delivery and has no bearing on database write latency.