Why Velox, ReadySet, and Neon Are Redefining the 2022 Database Landscape

1. A Rapidly Cooling Market

2021 was the most active year for database financing, with multi‑hundred‑million‑dollar rounds for Timescale, dbt Labs, Starburst, DataStax and SingleStore. Early 2022 continued the momentum, but macro‑economic headwinds halted later‑stage funding, leaving only smaller rounds such as Neon’s $30 M and MotherDuck’s $47.5 M, both backed by star‑studded founding teams. Domestic Chinese deals mirrored this pattern, with a record‑breaking angel round for SelectDB followed by a lull until Greptime’s modest raise near year‑end. MariaDB’s SPAC‑driven IPO and subsequent stock plunge illustrated the broader market contraction, yet technical innovation in the database space persisted.

2. Yearly Technology Highlight – Velox

Every database needs an execution engine. Historically each system built its own (e.g., Presto, Spark). Velox extracts common functionality into a reusable C++ library, offering components such as:

Type – a universal type system.

Vector – columnar, Apache Arrow‑compatible vectorization.

Expression Eval – expression evaluation leveraging vector capabilities.

Function – a pluggable function framework.

Operators – implementations of TableScan, Project, Filter, Aggregation, Order, Join, etc.

IO – integration with storage systems.

Resource Management – compute resource handling.

Implemented in C++, Velox is positioned as the STL of database engines: a high‑performance, engineering‑friendly foundation that can be adopted by new projects to add features like vector processing in weeks rather than months.

The typical query pipeline is Language Frontend → IR → Optimizer → Execution Engine → Execution Runtime . While SQL front‑ends have largely converged on PostgreSQL dialects, Velox aims to standardize the Execution Engine layer, leaving Optimizer and Runtime as the primary differentiators among engines.

Using the Formula 1 analogy, Velox is the chassis and drivetrain, allowing teams to focus on strategy (Optimizer) and tuning (Runtime) for different tracks (business scenarios).

3. Yearly Feature – ReadySet

ReadySet, founded by Jon Gjengset based on his MIT PhD thesis “Partial State in Dataflow‑Based Materialized Views”, provides a transparent cache layer that speaks the MySQL/PostgreSQL wire protocol. Applications see ReadySet as part of the database itself, eliminating the need for explicit cache‑write logic and addressing the classic cache‑invalidation problem for roughly 95 % of use cases.

The project is open‑source ( https://github.com/readysettech/readyset) and has inspired PlanetScale’s Boost feature. ReadySet exemplifies how academic research can be turned into a production‑grade database capability.

4. Yearly Database – Neon

Neon positions itself as the “PostgreSQL‑based PlanetScale”, offering a serverless, developer‑friendly workflow with a storage engine designed for branching. Its onboarding experience is extremely fast: creating a database and obtaining a connection string can be done in minutes.

Because Neon runs the native PostgreSQL server in the compute tier, it avoids compatibility issues that arise with distributed MySQL‑based solutions. While currently focused on OLTP workloads, its architecture leaves room for AP and time‑series extensions.

5. Other Notable Databases

Google AlloyDB – a PostgreSQL‑compatible, WAL‑based compute‑storage‑separated service that builds on the Aurora model but emphasizes AP capabilities.

Snowflake Unistore – an HTAP solution that extends Snowflake’s data‑lake roots into transactional workloads, highlighting the trend of “TP → AP” migration.

SQLite Ecosystem – 2022 saw SQLite’s WASM port become an official project, Cloudflare’s D1 serverless offering, LiteFS for distributed file‑system‑based replication, and DuckDB‑based services like MotherDuck, all pushing SQLite toward global, distributed use cases.

PostgreSQL – despite a slight drop in DB‑Engines ranking, it remains the most loved database on Stack Overflow and continues to dominate open‑source data platforms, with extensions such as Trusted Language Extensions (TLE) from AWS and widespread adoption in cloud PaaS offerings.

6. Emerging Trends

Fusion – converging SQL dialects, execution engines, and log‑based storage to enable a single framework that can serve TP, AP, and lake workloads.

AI‑driven tooling – while large‑language models can generate SQL from natural language, reliable optimizer‑level integration remains a challenge. The more practical path is to build databases that expose AI/ML capabilities via extended SQL syntax and platform services like DataZone.

Tooling ecosystem – the database tool market is expanding with acquisitions (MongoDB → Compass, Databricks → Redash, ClickHouse → Arctype) and new products (Bytebase, DataZone) aiming to provide comprehensive IDE‑style experiences.

7. 2022 Prediction Scoring

Retrospective scores were assigned to 2021 forecasts: PlanetScale’s growth was over‑hyped (50 % accurate), Neon and Xata’s developer‑workflow focus were spot‑on (100 % accurate), and Snowflake’s dominance was correctly predicted (0 % error). Overall prediction accuracy was 35.7 %.

8. 2023 Outlook

Key expectations include:

Microsoft integrating OpenAI models into SQL Server and Power BI.

Snowflake launching its own BI product and possibly acquiring a BI company.

Snowflake Unistore extending into data‑lake territory.

Fly.io delivering a globally distributed SQLite‑in‑the‑browser product.

DB‑Engine rankings: Snowflake, PostgreSQL, SQLite.

New databases targeting AI/ML workloads with PostgreSQL‑compatible protocols.

emergence of a heavyweight open‑source tool for database application development.

These trends suggest continued convergence of storage, compute, and analytics, with developer workflow and AI integration driving the next wave of innovation.