April 2025 Summaries

AlloyDB Omni is a PostgreSQL-compatible database that offers high compatibility with open-source PostgreSQL and provides additional features such as AI/ML workload integration, columnar data processing, and optimized memory management. Unlike other PostgreSQL-compatible databases, AlloyDB Omni's architecture is architecturally similar to PostgreSQL, providing the least vendor lock-in. However, its closed-source nature means that end-users have limited transparency into its internal workings. Aiven for AlloyDB Omni provides a managed solution for AlloyDB Omni, while Aiven for PostgreSQL offers more flexibility and data portability as an open-source solution. The choice between the two ultimately depends on the intended use case and requirements, such as AI/ML integration or traditional relational data storage.

The Hitchhiker’s guide to Diskless Kafka` proposes a new topic type for Apache Kafka that redirects replication from broker disks to cloud object storage, reducing the total cost of ownership (TCO) by up to 80 percent. This deep-dive into KIP-1150 explains how diskless topics eliminate cross-zone traffic and provide instant elasticity without changing the client or requiring costly migrations. The design is intentionally conservative, keeping the upgrade path straightforward with a single topic flag, while also introducing compaction agents to optimize object storage usage and reduce latency. The proposal aims to build on cloud primitives that already run half the internet, making it a significant improvement for Kafka users in the cloud.

Apache Kafka is introducing a new feature called Diskless Topics that replicates data directly in object storage, reducing the total cost of ownership (TCO) and improving performance. This new feature allows users to delegate replication to object storage, eliminating disks from the hot path and reducing costs by up to 80%. The Diskless Topic proposal aims to make Kafka more agile, scalable, and cost-effective, with features such as autoscaling in seconds, multi-region disaster recovery, and reduced IOPS bottlenecks. The new feature is designed to be backwards compatible with existing Kafka topics and can be used alongside traditional replicated topics. By leveraging cloud-native storage services like S3, Diskless Topics can provide a more efficient and cost-effective way to store and replicate data in the cloud.

Aiven for PostgreSQL has conducted a performance benchmarking comparison across four major cloud providers: AWS, GCP, OCI, and Azure. The benchmarks measured the transactions per second (TPS) of PostgreSQL instances on each provider, using similar configurations and varying database sizes. The results show that Aiven's managed PostgreSQL service outperforms the native services of the hyperscalers in terms of TPS and price. However, the performance difference is not always explicit in the overall pricing, making it essential to consider both factors when choosing a cloud provider for PostgreSQL. The benchmarks also highlighted the importance of considering latency and other network-related factors in evaluating database performance. Aiven plans to publish Terraform benchmarking scripts and extend the tests on a parameter settings level to provide more insights into performance improvements and comparisons with hyperscaler fork offerings.