Understanding Kafka Partitioning Strategy
Learned how Kafka uses
sticky partitioning to optimize batching. This significantly reduces latency when producers don't specify a key, as it keeps records pinned to a single partition until the batch is full.
Zero-copy Protobuf Serialization
Explored how to use
unsafe pointers in Go to achieve zero-copy deserialization for Protobuf messages. Great for high-throughput systems like Shadowstream.
LSM-Tree Compaction Strategies
Dived deep into Leveled vs Tiered compaction for GopherDB. Leveled compaction reduces read amplification but increases write amplification—a classic trade-off in storage engine design.