September 2026 Summaries

Stan, a Product Director at Aiven, draws on experience founding and fundraising for startups, as well as work in programming, marketing, PR, and events, to lead a newly created product area with a startup-style mindset. Recruited to build a team and product largely from scratch, he views his role as acting like the CEO of a small startup within a larger company, emphasizing autonomy, expert ownership, curiosity, and rapid decision-making. He joined Aiven after candid conversations with its leadership convinced him that its transparent culture and focus on difficult technical challenges aligned with his background and ambitions. While he values the scale and resources of Aiven, he also notes the adjustment required in navigating more process and managing limited context compared with the close-knit startups where he previously knew every employee.

Aiven’s Bring Your Own Key (BYOK) feature lets customers retain control of encryption keys in their own AWS KMS, Google Cloud KMS, or Azure Key Vault while granting Aiven narrowly scoped permissions to encrypt and decrypt service data. It is particularly relevant for regulated sectors, sovereignty requirements, and organizations seeking immediate revocation capabilities, since disabling or revoking the key can make data inaccessible without requiring provider intervention. Across all three clouds, implementation follows the same broad process of creating a key, granting Aiven access, registering it as a customer-managed key, and attaching it to a service, though cloud-specific IAM requirements can make GCP, AWS, and especially Azure setup more complex. Key rotation behavior differs by provider: AWS rotation is generally transparent, GCP requires retaining older key versions for existing encrypted data, and Azure may require updating Aiven’s key registration when version-specific key URIs are used. The feature transfers meaningful operational responsibility to customers, who must protect keys, maintain permissions, plan rotation and retention, and rehearse emergency revocation because Aiven cannot recover data made inaccessible through customer-side KMS changes.

Organizations using Snowflake or Databricks primarily for dashboards and scheduled reports may face costs that increase with user concurrency and continuously running compute, despite having relatively simple SQL-based reporting workloads. The passage proposes Aiven for ClickHouse as a fixed-plan, managed columnar database alternative for high-concurrency analytics, with integrations for PostgreSQL, Apache Kafka, cloud object storage, and DataHub intended to reduce the need for separate ingestion, orchestration, and governance tools. It describes using refreshable materialized views for smaller PostgreSQL tables, Kafka and Debezium for event streams or change data capture, and tiered storage to retain older data in lower-cost object storage while keeping it queryable. Rather than replacing Snowflake or Databricks entirely, it recommends migrating costly dashboards incrementally, validating results against existing reports, retaining the original platforms for large-scale transformations, Spark jobs, or machine learning, and shifting suitable dashboard traffic to ClickHouse.

Apache Kafka’s KRaft mode depends on a controller Raft quorum to manage cluster metadata, and losing a majority of controller voters halts control-plane operations because dynamic quorum membership can only be changed through records that require an existing quorum to commit. Aiven describes an emergency recovery process for such rare failures that cannot restore lost data or full consistency, but can restore availability by identifying the longest surviving metadata log across controllers and brokers, selecting it to avoid rolling surviving brokers back to earlier leadership epochs, and offline-editing it to make one controller the sole voter. That controller can then restart, become leader, allow other surviving nodes to replicate the revised membership state, and support replacement voters added through normal operations. The process deliberately discards metadata and potentially partition data that existed only on lost disks, so it is reserved for cases where failed voters are confirmed unrecoverable and is performed manually rather than automatically. The discussion also notes proposed upstream Kafka work on overriding stale voter endpoints, while emphasizing that this does not solve majority-loss recovery because lost voters must be removed from membership records, not merely assigned new addresses.

Aiven’s Model Context Protocol connector enables Base44’s AI agent to provision and manage Aiven PostgreSQL services, create and query tables, seed data, and deploy applications while keeping infrastructure in the user’s own Aiven account and chosen cloud region. After an Aiven organization administrator enables MCP access, users add the Aiven MCP endpoint in Base44 and authorize it through OAuth 2.0 with PKCE, avoiding the transfer of API keys or direct database credentials. The connector is intended for build-time use, with default redaction of service connection details and optional restrictions such as PostgreSQL-only scopes, read-only mode, and controlled write permissions. Base44 can then inspect available Aiven projects, plans, and regions, select and provision a suitable PostgreSQL service, prepare the data layer, and build the application. While Aiven’s free PostgreSQL tier supports provisioning and database setup, deployment requires a paid plan because service integrations are needed to pass credentials to the deployed app; users must add the database URI as a Base44 secret and select the Aiven integration during deployment.

OpenSearch User Behavior Insights (UBI), available by default on Aiven for OpenSearch, is presented as a way to connect search relevance work with real production behavior rather than relying solely on aging offline judgment sets. It stores search requests in the `ubi_queries` index and user actions such as clicks, scrolling, and dwell time in `ubi_events`, linking both through a shared `query_id` that preserves the relationship between ranked results and subsequent engagement. This data allows teams to identify high-volume queries with low engagement, distinguish ranking problems from weak result presentation, and use repeated clicks or dwell as implicit signals for curated rules, learning-to-rank models, and A/B tests. A demonstration application using product data illustrates how searches, engagement metrics, and underlying event records can be explored through a UI and OpenSearch Dashboards. While UBI does not replace expert evaluation, it helps relevance engineers, product teams, and catalog editors focus their investigation on observed user behavior, providing feedback that remains useful across traditional, vector, hybrid, and generative search systems.

Aiven Runtime, formerly Aiven Apps, is now generally available as a platform for deploying containerized applications and AI agents alongside managed data services, addressing the infrastructure, security, and networking challenges that often arise when moving local projects into production. It supports standard Containerfiles and Docker Compose configurations, automatically builds and deploys applications, detects compatible services such as PostgreSQL, MySQL, and Valkey, and connects them without requiring separate VPC, IAM, secret-management, or egress setup. Available across AWS and Google Cloud, with Azure planned, the service is intended to let developers run applications, monitoring tools, internal APIs, data pipelines, and production AI agents close to governed live data, reducing latency and exposure risks. Runtime is part of the Aiven Data & AI Cloud, alongside managed open-source technologies including Kafka, PostgreSQL, ClickHouse, OpenSearch, Valkey, and DataHub, while Aiven MCP is positioned as a secure standard interface for agents to access approved data.

Aiven presents its Data & AI Cloud as a managed, open-source data platform designed to support the growing number of developers, business teams, applications, and AI agents building on organizational data. It argues that fragmented stacks of separate vendors or self-managed open-source components create integration, operational, governance, latency, and lock-in challenges that can prevent AI systems from safely using data. The platform aims to provide fresh, low-latency, governed, accessible, and open data through managed versions of Apache Kafka, PostgreSQL, ClickHouse, OpenSearch, and Valkey, alongside Aiven DataHub for cataloging, lineage, governance, and MCP-based agent access, and Aiven Runtime for running AI applications near data. Aiven emphasizes a unified control plane, enterprise security and compliance features, a 99.99% uptime SLA, deployment across AWS, Google Cloud, Azure, and bring-your-own-cloud environments, and portability based on upstream, unmodified open-source technologies and open standards.

Aiven has announced the general availability of Aiven DataHub, a fully managed open-source data catalog designed to help organizations and AI agents discover, understand, govern, and trust data across distributed systems. The service addresses problems such as inconsistent definitions, unclear ownership, fragmented data sources, and exposure of sensitive information by providing searchable metadata, ownership details, classification, governance controls, and column-level lineage. Early users, including UK payments provider Dojo, report faster data discovery, reduced reliance on engineers for locating datasets, less duplication, improved visibility into obsolete systems, and more reliable AI-assisted work through DataHub’s MCP server. Built on the open-source DataHub core, the offering has no per-seat licensing and supports unlimited users and agents, aiming to encourage organization-wide adoption. General availability adds a 99.9% uptime SLA, managed security patches, upgrades, scaling, cloud migration support, and performance claims involving sub-100-millisecond entity lookups during concurrent user, agent, and ingestion workloads. DataHub supports more than 100 connectors for sources including BigQuery, Snowflake, dbt, MongoDB, and databases, while offering simplified integration and automatic lineage for Aiven services such as Kafka, PostgreSQL, ClickHouse, OpenSearch, and MySQL.

Katja Rantala, VP of People at Aiven, describes her career as an unplanned shift from studying international politics and expecting a diplomatic career to discovering a talent for recruitment and people leadership in high-growth technology companies. Drawing on experience guiding international businesses through rapid growth, restructuring, and strategic change, she emphasizes ambition, creative problem-solving, and accepting imperfection rather than pursuing unattainable perfection in volatile environments. At Aiven, she supports a small, hands-on organization where leaders remain closely involved in practical work, employees are expected to be self-directed, and in-person interaction remains valued despite the company’s broader flexibility. Rantala says Aiven is not suited to everyone because it is still evolving and tackling difficult technical and business challenges, but she views its current effort to reinvent its strategy and operations as an opportunity for employees to make a meaningful impact.