On March 12th & 13th, 2026, I had the privilege of staffing the Aiven booth at PGConf India in Bengaluru, one of the most energetic database conferences in the Asia-Pacific region. Alongside Pranay, Padma and the OpenSource DB team, we spent the day deep in conversation with DBAs, platform engineers, and architects who live and breathe data infrastructure. What struck me most wasn’t just the volume of footfall, it was the quality of questions.This post is a distillation of what we discussed, what confused people, and what ultimately resonated.
The conversations that mattered
PGConf India draws a technically sharp crowd, and the questions at our booth reflected that. Attendees weren’t asking what PostgreSQL is. They were asking how Aiven handles synchronous vs. asynchronous replication under real failover conditions, whether logical decoding to Kafka held up under transactional load, and, most persistently, how Aiven’s operational model actually works inside their cloud environment.
Three technical themes dominated the day.
Theme 1: Replication and high availability, the details matter
The HA story is table stakes at any Postgres conference. What engineers here wanted to dig into were the operational specifics: what happens during a failover, who orchestrates it, and how visible is the topology to the application layer.
Aiven manages PostgreSQL replication in both synchronous and asynchronous modes, with automated failover built in. Cascading replication, where secondaries replicate from other secondaries rather than the primary, is supported and handled transparently by Aiven’s orchestration layer, so customers don’t have to manage topology by hand. Read replicas are exposed as first-class endpoints for horizontal read scaling and workload offloading.
For teams running high-traffic applications that cannot afford to hand-roll failover logic, this abstraction is significant. The question is less “can we do this” and more “do we trust the platform to do it correctly under pressure?”, a fair concern, and one we addressed with real deployment scenarios from production environments.
Theme 2: Kafka + PostgreSQL CDC, the event-driven data layer
Change Data Capture pipelines from PostgreSQL to Kafka generated some of the most animated discussions at the booth. This architecture, using PostgreSQL’s logical decoding to stream row-level changes into Kafka topics in near real-time, has become a cornerstone pattern for teams decoupling their application and analytics layers.
The value proposition is clear: your database stops being a bottleneck for downstream consumers. Microservices, analytics pipelines, and data lake ingestion jobs can all subscribe to a reliable, ordered event stream without polling the primary database or coupling themselves to its schema lifecycle.
Aiven’s managed Kafka offering handles the Kafka side of this equation, while the Aiven connector ecosystem bridges it to PostgreSQL’s logical replication slots. For teams already running managed Postgres on Aiven, the integration story is relatively tight. For those bringing their own Kafka or Postgres, there’s more integration work involved, and we were upfront about that trade-off.
Theme 3: What Aiven actually is, clearing up real confusion
This one surprised me the most, and I want to document it honestly because it came up repeatedly.
A significant number of visitors initially assumed one of two things: either that Aiven builds its own fork of PostgreSQL (it doesn’t), or that adopting Aiven means somehow “installing Aiven into your cloud” like a self-hosted tool (also not how it works).
Let me be direct on both points.
Aiven does not develop PostgreSQL. It is a managed service provider that runs standard, upstream PostgreSQL on your behalf. The value isn’t in a modified database, it’s in everything around the database: automated provisioning, backups, scaling, monitoring, high availability, and day-two operations that most engineering teams don’t want to build from scratch.
BYOC (Bring Your Own Cloud) is where the second misconception clusters. In Aiven’s standard model, Aiven runs your services in its own managed cloud accounts, you connect to an endpoint, and the underlying infrastructure is Aiven’s problem. With BYOC, the infrastructure runs inside your own AWS, GCP, or Azure account, but Aiven still manages everything operationally. You own the billing relationship, the networking boundary, and the security perimeter. Aiven handles deployment, scaling, updates, and monitoring.
What BYOC is not is a self-hosted installation. You don’t configure Aiven tooling inside your cloud. You connect your cloud account to Aiven’s control plane, and Aiven uses that access to provision and manage resources on your behalf. This is a subtle but important distinction, and getting it wrong leads to significant integration planning errors.
The broader platform picture
Beyond PostgreSQL, Aiven’s managed portfolio spans Apache Kafka, OpenSearch, MySQL, and Redis, all cloud-agnostic and deployable across major hyperscalers. The observability layer is worth calling out specifically: CPU, memory, I/O, query latency, and replication lag are all surfaced through the Aiven Console, with integrations for Prometheus and Grafana for teams running their own monitoring stacks. Scaling, both vertical (resizing compute and storage) and horizontal (read replicas, Kafka partitioning), is designed to happen with minimal downtime.
For teams evaluating a managed data platform, the core question is always whether the operational simplification justifies the trade-off in control and cost. At the booth, the engineers who were most excited were those already drowning in database operations, the ones who know exactly how painful it is to manage HA, backups, and scaling by hand. For them, Aiven’s value proposition landed immediately.
Reflections from the floor
Staffing a booth at a technical conference is a unique experience. You get an unfiltered view of what practitioners actually struggle with, not what they say in surveys or RFPs, but what they ask when they have five minutes and a real problem. The recurring misconceptions around Aiven’s operational model were a useful reminder that even experienced engineers approach managed services with mental models shaped by self-hosted tools, and that’s a communication challenge worth taking seriously.
PGConf India continues to be one of the most substantive regional database conferences in the world. The conversations we had reaffirmed why community engagement, not just marketing, matters in this space.
If you have questions about Aiven’s architecture, BYOC model, or how CDC pipelines work in practice, feel free to reach out or continue the conversation in the comments.Thanks to everyone who stopped by the Aiven booth.