July 2023 Summaries

Neon is a fully managed serverless Postgres service that allows users to create isolated copies of their database, referred to as "branches", using a process called "branching". This feature enables users to programmatically restore their data to a previous state by creating a branch from a specific point in time. Neon retains a data history in the form of Write-Ahead-Log (WAL) records, allowing for restoration to a previous point in time within a configurable seven-day window. Users can create branches using the Neon CLI or console, and restore them to a previous state while keeping the same compute endpoint by assigning it to the newly created branch and setting it as the primary branch.

Neon's branching feature offers a flexible solution for instant data recovery by allowing users to restore data to a previous state in a serverless Postgres environment. This is achieved through creating isolated copies of the Postgres cluster, known as branches, using a copy-on-write method that is both fast and cost-effective. Branches can be created from any point within a seven-day data history window, leveraging Write-Ahead Log (WAL) records to ensure data integrity. The feature also supports maintaining the same compute endpoint during data restoration, eliminating the need for connection string changes. This process can be managed programmatically via the Neon CLI or API, and a TypeScript script example is provided to guide users in implementing this functionality. Users are advised to delete or rename old branches to manage storage costs effectively.

Exploring the innovative concept of database branching with Neon, the article demonstrates how developers can utilize this feature to create a Discord bot for automating coding events and contest fulfillment. Neon offers a unique branching model similar to git, allowing inexpensive, copy-on-write operations that provide flexibility for diverging data structures and facilitating point-in-time recovery. The guide walks through setting up a Discord bot that gathers shipping information for prize fulfillment using Neon branches, with a focus on the practical application of creating, managing, and deleting branches as needed. The use of Neon’s branching is highlighted as a strategy to manage disposable branches for temporary data, with integration into the development workflow through tools like PostgresJS, Prisma, and Harmony for Discord, leveraging Neon’s REST API and CLI tools for seamless integration into CI/CD processes. The article underscores the advantages of Neon’s branching for enhancing development workflows, particularly in startups, by enabling scalable and efficient data management and backup strategies.

The Neon team has made significant improvements in reducing the "cold start time" for compute resources, which refers to the time it takes for idle computes to become active again. The current cold start time is around 500 milliseconds, a major improvement from previous times of up to 3-6 seconds. This achievement was made possible by applying optimization strategies such as only reconfiguring when necessary and creating a concept of "compute pools" that always have an available compute ready for use. Additionally, the team has implemented changes in configuration tracking, networking, caching, concurrency, and autoscaling to further improve performance. The ultimate goal is to reduce the cold start time even further, with hopes of reaching 50 milliseconds or less.

Neon's engineering team has significantly reduced the cold start time for compute resources, cutting it down from an average of 3-6 seconds to around 500 milliseconds. This improvement was achieved by re-evaluating the problem and applying two main optimization strategies: reconfiguring only when necessary and utilizing compute pools. The first strategy involves avoiding unnecessary configuration updates unless a compute is idle and requires a health check, while the second strategy focuses on having pre-started computes ready to be configured on demand, thus bypassing the longer start times associated with on-demand setups. Additionally, improvements in networking, caching, and concurrency, along with a new configuration system, have further decreased start times. The ongoing efforts include autoscaling and optimizing resource allocation to maintain efficiency, with the ultimate goal of achieving consistent speeds across all regions. These enhancements are part of a broader initiative to improve Neon’s serverless offerings by making cold starts a thing of the past and ensuring faster and more efficient compute resource management.

Our company has achieved SOC2 Type 1 compliance, a significant milestone in its commitment to security. SOC2 is an auditing standard developed by the American Institute of Certified Public Accountants (AICPA) that ensures service organizations' non-financial controls related to security, availability, processing integrity, confidentiality, and privacy are robust and effective. The company pursued SOC2 compliance to demonstrate trust with customers and partners, improve its systems and controls, and differentiate itself in the market. Achieving SOC2 compliance involved a rigorous process of gap analysis, policy and procedure development, control implementation, training and education, and an independent audit. To address background check concerns, the company implemented a comprehensive reference check policy. Now that it has achieved SOC2 Type 1 compliance, the company plans to maintain its current level of compliance, work towards SOC2 Type 2 compliance, and continue innovating and improving its security measures based on user feedback and best practices.

Neon has achieved SOC2 Type 1 audit compliance, marking a significant step in its commitment to data security and privacy. SOC2 is an auditing standard by the American Institute of Certified Public Accountants (AICPA) that evaluates non-financial controls related to security, availability, processing integrity, confidentiality, and privacy. Neon pursued compliance to build trust, improve its systems, and differentiate itself in the market. The process involved a thorough gap analysis, policy and procedure development, control implementation, and training, culminating in a successful audit by an independent external auditor. Due to global legal variations, Neon opted for a comprehensive reference check policy instead of traditional background checks, balancing diligence with legal compliance. Following this achievement, Neon plans to maintain its compliance and work towards SOC2 Type 2, which requires demonstrating sustained compliance over time, while continuing to enhance its security measures based on best practices and user feedback.

The latest improvements made to ensure a better developer experience when using Prisma with Neon include reducing cold starts, having a better experience working with Prisma Migrate, and improving connection pooling support. These enhancements aim to provide a frictionless experience for Prisma users.

Neon has implemented several improvements to enhance the developer experience for Prisma users, focusing on reducing cold start times and providing better support for Prisma Migrate and connection pooling. By automatically scaling compute resources, Neon minimizes resource overprovisioning costs, but this scaling can lead to cold starts, which have been addressed by adjusting connection timeout settings. Additionally, Neon's integration with PgBouncer allows for efficient connection pooling, supporting up to 10,000 concurrent connections. To facilitate seamless development and production workflows, both pooled and direct database connections can now be used simultaneously, catering to the specific needs of Prisma Migrate and Prisma Client. Moreover, new features allow for automatic management of shadow databases, eliminating the need for manual creation and improving the initial setup process for developers using Prisma with Neon. These updates are part of Neon's ongoing efforts to optimize the cloud-based Postgres experience and invite user feedback for further enhancements.

With the release of Neon CLI, developers can now manage Neon resources directly from their terminal, streamlining database management with a command-line interface that handles authentication and manages projects, branches, databases, roles, and more. The CLI offers an integrated approach to managing various aspects of projects, including CRUD operations on projects, branches, databases, and roles, allowing users to create Postgres databases, list branches, and create new ones for local development. Additionally, the CLI can be integrated into CI/CD pipelines and used with a NEON_API_KEY to manage Neon resources efficiently. The development team is actively working on adding more features to the CLI, encouraging feedback from users to improve their experience managing Postgres databases.

Neon has introduced Neon CLI, a command-line interface that enables developers to manage Neon resources directly from the terminal, enhancing the efficiency of database management for Postgres. The CLI facilitates a wide range of operations, including authentication, project and branch management, and CRUD operations on databases and roles, all without leaving the command line. It allows users to create Postgres databases and branches quickly, providing connection strings immediately for seamless integration into projects. The CLI also supports integration into CI/CD pipelines, requiring a NEON_API_KEY for authentication, and offers flexible management of isolated database branches for development and testing. Neon is actively developing the CLI and encourages user feedback to refine its features, inviting users to explore its capabilities and contribute to improving the management of Postgres databases.

The `pg_embedding` extension for Postgres and LangChain offers a 20x speed boost for graph-based approximate nearest neighbor search while maintaining 99% accuracy, outperforming the popular IVFFlat indexing technique used in pgvector. The new extension uses the Hierarchical Navigable Small Worlds (HNSW) index to unlock new levels of efficiency in high-dimensional similarity search. pg_embedding can be easily implemented in applications, and prior knowledge of vector indexes is optional. Its performance is comparable to IVFFlat but with better speed and accuracy. The choice between `pg_embedding` and pgvector with IVFFlat depends on specific use cases and requirements, including memory constraints, search speed, accuracy, distance metrics, and index construction speed.

The text compares the performance of two protocols, SQL-over-HTTP and WebSockets, for executing Postgres queries in Edge and Serverless environments. The author's driver initially supported only WebSockets but recently introduced SQL-over-HTTP to compare their performance. The results show that while WebSockets are faster after establishing a connection, they are slower for single-shot queries compared to SQL-over-HTTP. However, when using HTTP with connection caching, latencies can be reduced further. The choice of protocol depends on the query type and region, as well as the Edge network's size and proximity to users.

Neon introduces same-region read replicas to its serverless architecture, enhancing database efficiency and scalability. This feature is available on Neon's Pro plan and can be accessed through the console or CLI. Read replicas in PostgreSQL are used for scaling read operations, reporting, analytics, and disaster recovery. Neon's read replicas separate storage and compute, offering efficient storage, data consistency, scalability, cost-effectiveness, resource customization, and access control.

In the exploration of optimizing SQL query execution at the edge, a comparative analysis between SQL-over-HTTP and WebSockets reveals distinct advantages for each protocol. The Neon driver, initially designed to facilitate Postgres database queries via WebSockets, encountered challenges with network round-trips in edge environments, which were mitigated through optimizations that reduced connection latency significantly. Experiments demonstrated that while WebSockets excel in multi-query scenarios once a connection is established, SQL-over-HTTP offers lower latency for single-shot queries due to its compatibility with connection caching. The study highlights that the choice between these protocols depends on specific use cases and runtime environments, as Edge networks provide geographic proximity advantages while serverless environments with connection caching can yield even lower latencies. Developers are encouraged to consider query types, geographical user distribution, and specific API requirements when selecting a protocol to ensure optimal performance in their applications.

The release of the pg_embedding extension for Postgres and LangChain introduces a significant enhancement in vector search capabilities, offering a 20-fold increase in speed with 99% accuracy using the Hierarchical Navigable Small Worlds (HNSW) index for approximate nearest neighbor search. While the existing pgvector extension with IVFFlat indexing has been popular, pg_embedding's graph-based approach in high-dimensional similarity search demonstrates superior efficiency. Benchmark tests using the GIST-960 Euclidean dataset show that pg_embedding with HNSW outperforms pgvector with IVFFlat in terms of search speed and accuracy, though it requires more memory due to its graph structure. This makes pg_embedding an ideal choice for applications prioritizing search speed and accuracy, while pgvector remains suitable for scenarios with strict memory constraints. Ultimately, the choice between the two should be guided by the specific requirements of the user’s application, with pg_embedding offering a powerful tool for efficient vector similarity searches in databases.

Neon has introduced same-region read replicas to its serverless Postgres architecture, enhancing database efficiency and scalability. These read replicas allow the primary database to focus on write operations by offloading read queries, thereby improving performance. Unlike traditional PostgreSQL read replicas that duplicate data across instances, Neon's architecture separates storage and compute, streaming data from a single source to all replicas, ensuring consistency and reducing storage costs. This approach not only supports higher throughput and workload offloading but also offers resource customization and access control. Neon's features, such as autoscaling and the ability to scale read replicas independently, further optimize resource management, making it a cost-effective solution for managing varying workloads. The introduction of these read replicas is part of Neon's broader strategy to achieve global replication, promising a robust and scalable solution for PostgreSQL databases.

We are excited to announce the release of our enhanced Postgres query driver for Vercel Edge Functions, which reduces query latencies by 40% and brings same-region queries down to single-digit milliseconds, providing a faster user experience. The driver is optimized for one-shot queries and connection caching, and is backward-compatible with existing Neon and Vercel Postgres code. We recommend using the new syntax with neon for single queries, while creating a new connection with Pool for multiple queries.

Vercel has introduced an enhanced driver for Postgres queries, significantly reducing query latencies by 40% for Vercel Serverless and Edge Functions, bringing same-region queries to sub-10ms levels and improving user experiences. This performance boost is achieved through SQL-over-HTTP via fetch support and experimental connection caching with Neon's proxy, suitable for single query executions where connection establishment times traditionally increase latency. The new driver is compatible with existing Neon and Vercel Postgres code, offering developers guidance on optimizing latency with single or multiple queries by choosing between the new syntax with neon for single queries or Pool for multiple queries. As edge computing evolves, Vercel is committed to further latency improvements, encouraging developers to experiment with Neon and provide feedback through community forums.

The Neon Partner Program has been launched to enable seamless integration of Neon into any application. It offers two types of integrations - OAuth and API. OAuth provides a secure method for third-party applications to obtain user data without sharing login credentials, while the API allows partners to create serverless Postgres databases effortlessly. Several partners have already integrated Neon into their products, including Cloudflare, Hasura, WunderGraph, Vercel, and Replit. By partnering with Neon, developers can easily add Postgres to their product while offloading the work of managing and scaling infrastructure.

Neon has launched its official Partner Program, enabling seamless integration of its database technology into various applications. The program offers OAuth and API integrations, allowing partners to securely access and manage Neon resources without sharing user credentials. Key partners, including Cloudflare, Hasura, and WunderGraph, utilize these integrations to enhance their platforms by incorporating Neon databases, thereby facilitating serverless execution and streamlined API development. Additionally, companies like Vercel and Replit have integrated Neon Postgres to simplify the development and deployment process for database-backed applications. Partnering with Neon offers the advantage of incorporating Postgres databases into products while benefiting from expert support and efficient infrastructure management, making it an attractive proposition for developers and companies looking to enhance their technological offerings.