Description: Currently we have batch tuning and Transactional offload tuning are set to default values for every new task. We initially perform a tuning exercise to meet latency SLA’s and set the task memory/buffers accordingly. Over time, month/quarter end loads, we see latency spike up because of a long running or huge transaction or sudden influx of data and we miss our latency SLA’s.
Ask:
- Add intelligence in the product to update the tuning settings automatically by reading source volume/velocity patterns.
- Additionally consider the overall system limitations when performing the intelligent tuning.
Outcome: will help in minimizing the number of times SLA’s are missed and avoid performance tuning exercises for tasks every now and then.