If we store transaction details in Kafka and consume them later, even with partitioning, is that enough? How many resources will this consume?

Kafka is designed for high‑throughput, low‑latency messaging. The resources consumed depend on several factors: the number of partitions, replication factor, message size, and consumer group configuration. Each partition is a log file; more partitions increase parallelism but also require more disk I/O and CPU for producers and consumers. A replication factor of 3 means each message is written to three brokers, increasing network traffic and disk usage. Consumers read sequentially from partitions; a consumer group with one consumer per partition can achieve maximum throughput. CPU usage is modest for simple string messages but rises with compression or serialization overhead. Disk usage is linear with the retention period and message size. For a typical transaction log of 1 MB per message, 10 k messages per second, 3 replicas, and 100 partitions, you’d expect ~30 GB/s of write traffic, ~10 GB/s of read traffic, and several hundred MB of disk per day. Monitoring tools (Kafka Manager, Prometheus) help tune these parameters. In summary, Kafka can handle millions of messages per second with modest resources if partitions and replication are tuned appropriately.