Accurate Azure Disk TX Cost
Due to the way we calculate IO cost by assuming the worst case, it can result in extremely inaccurate disk costs in Azure (25%). And since disk costs dominate Azure Kafka costs, it can directly result in a deployment cost that’s up to 25% inaccurate! (🚨🚨🚨)
2minutestreaming About 1 year ago
Accurate Azure Disk TX Cost
Due to the way we calculate IO cost by assuming the worst case, it can result in extremely inaccurate disk costs in Azure (25%). And since disk costs dominate Azure Kafka costs, it can directly result in a deployment cost that’s up to 25% inaccurate! (🚨🚨🚨)
2minutestreaming About 1 year ago
Account for Zookeeper / KRaft nodes
The calculator currently doesn’t account for KRaft/ZooKeeper nodes. This issue tracks adding support for it. - What instance should it be? - how many - etc
2minutestreaming About 1 year ago
Account for Zookeeper / KRaft nodes
The calculator currently doesn’t account for KRaft/ZooKeeper nodes. This issue tracks adding support for it. - What instance should it be? - how many - etc
2minutestreaming About 1 year ago
Cluster Utilization Buffer / Performance Buffer / Avg-Max Throughput / Latency Profile
As mentioned in the docs, the calculator could use one of the following: high-level: latency profiles: some sort of pre-made latency profiles which chooses "low latency < 100ms", med latency "< 250ms p99" and etc. Then, the calculator chooses the right disk and buffer based on that. avg-max throughput: allow users to define their avg throughput for cost calculating purposes (e.g network charges) but configure a max that the cluster must be able to support without trouble at all times low-level: an IOPS utilization buffer. e.g ensure we only use 50% of provisioned IOPS
2minutestreaming About 1 year ago
Cluster Utilization Buffer / Performance Buffer / Avg-Max Throughput / Latency Profile
As mentioned in the docs, the calculator could use one of the following: high-level: latency profiles: some sort of pre-made latency profiles which chooses "low latency < 100ms", med latency "< 250ms p99" and etc. Then, the calculator chooses the right disk and buffer based on that. avg-max throughput: allow users to define their avg throughput for cost calculating purposes (e.g network charges) but configure a max that the cluster must be able to support without trouble at all times low-level: an IOPS utilization buffer. e.g ensure we only use 50% of provisioned IOPS
2minutestreaming About 1 year ago