Introducing AlloyDB, a PostgreSQL-compatible cloud database service

Introduction

In this section, Andi Guttmans introduces herself and shares the exciting product announcement that will be discussed in the session.

Andi Guttmans Introduces Herself

• Andi Guttmans introduces herself as the General Manager and VP of Engineering for Databases for Google Cloud.

Exciting Product Announcement

• Andi Guttmans announces an exciting product announcement to be shared in this session.

The Importance of Cloud Databases

In this section, Andi Guttmans discusses the importance of cloud databases and how they offer organizations agility, cost savings, and differentiated capabilities.

Importance of Cloud Databases

• The cloud offers organizations agility, cost savings, and differentiated capabilities.

• 75% of all databases are expected to be in the cloud this year.

• New applications are already being built in a cloud-first manner.

• Existing applications are moving to the cloud at a rapid pace.

Google Cloud Database Services

• Google Cloud offers a full suite of database services including Memorystore for Redis, Cloud SQL, Spanner, Firestore and Bigtable.

• These services help manage the complexity of running your databases so that they are secure, simple to use and operate making app development faster and easier for you.

Scale of Google's Databases

In this section, Andi Guttmans discusses the scale at which Google's databases operate.

Scale of Spanner Database

• Spanner processes over 2 billion requests per second at peak.

Scale of Bigtable Database

• Bigtable processes over 5 billion requests per second at peak.

Scale of Firestore Database

• Firestore powers applications with more than 750 million monthly active end users.

Postgres and Open Source Databases

In this section, Andi Guttmans discusses the benefits of open source databases, specifically Postgres.

Benefits of Postgres

• Postgres is an open source database that provides portability and a license-free cost model.

• It supports parallel query, partition tables, stored procedures, replication, change data capture, rich data models and more.

• Postgres's ecosystem of extensions helps it address a wide variety of use cases.

• It plugs into every major application development framework or ecosystem.

Popularity of Postgres

• Postgres is adopted by millions of users in every industry from startups to large enterprises.

• Despite the benefits, there's too much burden on developers to make it run well in production.

Conclusion

In this section, Andi Guttmans concludes the session by discussing how Google Cloud is working to support customers moving to the cloud with their applications as is.

Supporting Customers Moving to the Cloud

• Google Cloud is working hard to support customers coming from legacy proprietary databases.

• These customers are increasingly standardizing on open source and in particular one open source database - Postgres.

• There's still work to do to support these customers moving to the cloud with their applications as is.

Introduction to AlloyDB

In this section, the speaker introduces AlloyDB, a new database service that combines Google's cloud-native architectures with full Postgres compatibility. The speaker explains how AlloyDB is designed to help customers solve operational data problems and provides industry-leading performance and availability.

Key Points:

• Google started on the journey with Postgres to address user pain points.

• AlloyDB is a new database service that combines Google's cloud-native architectures with full Postgres compatibility.

• AlloyDB provides industry-leading performance and availability for top-tier relational database workloads.

• Full Postgres compatibility means regaining the freedom to run applications when and where you want and benefiting from the entire Postgres ecosystem.

Transactional Workloads

In this section, the speaker discusses how AlloyDB can handle transactional workloads better than standard Postgres or Amazon's comparable Postgres-compatible service. The speaker explains how AlloyDB delivers four times faster transactional workloads than standard Postgres and more than two times faster than Amazon's comparable service.

Key Points:

• AlloyDB is four times faster than standard Postgres for transactional workloads.

• Query performance scales linearly as you add more compute resources.

• Customers use operational databases for a variety of workloads, including real-time insights and reporting.

Disaggregating Compute and Storage

In this section, the speaker explains how Google's innovative approach around completely disaggregating compute and storage enables them to build truly unique systems. The speaker describes how they have built a new intelligent database optimized storage layer that scales separately from the database instances and offloads work from the primary.

Key Points:

• At AlloyDB's core is Google's innovative approach around completely disaggregating compute and storage.

• For AlloyDB, they have built a new intelligent database optimized storage layer that scales separately from the database instances and offloads work from the primary.

• The storage engine itself has disaggregated compute and storage, allowing them to transparently rebalance load if there's a hotspot.

• The storage service isn't the only technology powering fast, predictable performance. AlloyDB features automatic data tiering between DRAM, ultrafast cache, and durable storage.

Conclusion

In this section, the speaker concludes by summarizing how AlloyDB delivers industry-leading performance and availability for top-tier relational database workloads. The speaker emphasizes how AlloyDB is designed to help customers solve operational data problems with real-time insights and reporting.

Key Points:

• AlloyDB delivers industry-leading performance and availability for top-tier relational database workloads.

• Customers can use AlloyDB to solve operational data problems with real-time insights and reporting.

AlloyDB: A High-Performance, Cloud-Native SQL Database

The speaker introduces AlloyDB, a cloud-native SQL database that leverages modern processors to speed up query execution. The architecture allows for the best performance for both transactional and analytical workloads.

Introduction to AlloyDB

• AlloyDB is a cloud-native SQL database that uses modern processors to speed up query execution.

• The architecture of AlloyDB allows for the best performance for both transactional and analytical workloads.

• You can run super fast, real-time queries on your operational data directly with zero impact on your operational workload-- no schema changes, no additional indexes, no additional configuration.

• AlloyDB delivers real-time insights out of the box.

Call Center Use Case

• A call center CRM database is used as an example to demonstrate how much faster AlloyDB is than PostgreSQL.

• Four concurrent queries are run looking for the average customer's hold time during July and August 2021, total number of clients who called during those months, client that had maximum hold time, and the time of peak call volume.

• On the left side of a dashboard we have an AlloyDB instance while on the right side we have a PostgreSQL instance. Results show that running these four queries on PostgreSQL took an average of 164 seconds while running them on an AlloyDB instance they completed in less than a second.

Reliability and Availability

• The architecture enables a new level of reliability and availability with four-nines availability SLA that not only includes unplanned failures but also database maintenance as well.

• Recovery from most database failures within 60 seconds is independent of database size and load.

• Other reliability issues like out-of-memory errors and out-of-disk errors are a thing of the past.

• Writes on AlloyDB scale linearly with core count up to the largest instance supported on Google Compute Engine. Reads scale horizontally, with loads balancing across more than a thousand vCPUs.

Scaling Read Connections with AlloyDB

Talip Ozturk, Engineering Manager for AlloyDB read pools, demonstrates how to increase read performance by horizontally scaling read connections through adding more nodes.

Increasing Read Performance

• To double the read capacity on a cluster, edit the pool and change the node count from 2 to 4.

• Multiple clusters can be created with different node counts that are all backed by the same primary and database storage.

• Query per second should double from one node to two nodes and two nodes to four nodes.

• Running a standard pgbench select-only workload for 10 seconds against each pool shows an increase in transactions per second from 71,000 to 149,000 to 300,000.

Handling Primary Failure

• Even if the primary is down, read pools can continue running just fine.

AlloyDB Features and Integration with AI/ML Services

Andi Gutmans discusses how AlloyDB integrates with Google's portfolio of AI/ML services and offers autopilot systems for comprehensive database monitoring, troubleshooting, and optimizations.

Autopilot Systems

• AlloyDB features autopilot systems that automatically manage and dynamically adapt database systems to customer workload needs.

• Memory management, vacuuming, data tiering are complex tasks that AlloyDB automatically addresses.

Integration with AI/ML Services

• Seamless integration with Google's portfolio of distinctive AI/ML services such as Vertex AI for easy inferencing with AlloyDB databases.

• Pre-integrated with Vertex AI, developers can create a model in Vertex AI and then call it within a query or transaction.

• High-throughput, low-latency augmented transactions without having to write any additional application code.

Postgres Compatibility

• Fully compatible with Postgres including broad extensions and flag support.

• Launching with Postgres 14 compatibility but will have cadence with the latest releases.

• Move your Postgres application as is with no need for application changes using pg_dump and pg_restore or use their Database Migration Service.

Introduction to AlloyDB

In this section, the speaker introduces AlloyDB and explains how it helps customers modernize faster and reduce costs. The speaker also highlights the Postgres compatibility, scale, and performance of AlloyDB.

What is AlloyDB?

• AlloyDB is a fully Postgres-compatible database that allows customers to modernize faster and reduce costs.

• Leaders love it because it enables them to modernize existing workloads in the cloud, solve scaling issues, reduce management overhead, or enable real-time insights.

• Developers love the Postgres compatibility, along with the scale and performance of AlloyDB.

Three Fully Managed Services for Postgres Workloads

In this section, the speaker explains that Google Cloud offers three fully managed services for Postgres workloads: Cloud SQL, AlloyDB, and Spanner.

Three Fully Managed Services for Postgres Workloads

• Google Cloud offers three fully managed services for Postgres workloads: Cloud SQL, AlloyDB, and Spanner.

• Cloud SQL is an enterprise-ready relational database service for Postgres that provides strong management capabilities.

• AlloyDB is a fully Postgres-compatible database ready for top-tier workloads and legacy proprietary databases in the cloud.

• Spanner is a transformative relational database service with unlimited global scale.

Oracle-to-Postgres Migration Support

In this section, the speaker discusses Google Cloud's recent announcement of Oracle-to-Postgres schema and data migration support in Database Migration Service.

Oracle-to-Postgres Migration Support

• Google Cloud recently announced Oracle-to-Postgres schema and data migration support in Database Migration Service.

• Database Migration Service integrates with the Ora2Pg open source tool for schema conversion and provides serverless data movement and monitoring for migrations.

• Database Migration Service will also support easy upgrades from Cloud SQL to AlloyDB, so customers can run their Postgres workloads on Cloud SQL now and later move them easily to AlloyDB.

• Google Cloud has invested in a comprehensive end-to-end program to help make migration experience fast, easy, and cost-effective.

Getting Started with AlloyDB

In this section, the speaker explains how to get started with AlloyDB.

How to Get Started with AlloyDB

• To get started with AlloyDB, customers can create a cluster using Google Cloud's UI, API or gcloud CLI tool.

• Customers coming from Postgres can use pg_dump and pg_restore to get some data into AlloyDB.

• The preview of AlloyDB is currently being offered for a limited time at no charge.