Interview with Google Cloud
Any views or opinions represented or expressed in this interview belong solely to the interviewee and do not neccessarily represent those of the PGConf.EU 2022 organization, PostgreSQL Europe, or the wider PostgreSQL community, unless explicitly stated.
- In which areas do you expect PostgreSQL to grow most and how does your company contribute to and benefit from that growth?
As we work with our customers to modernize their IT infrastructure and applications, we see two important trends with respect to database management systems (DBMS).
The first is the migration of on prem applications and the DBMS they run on to the cloud in order to achieve higher agility, scalability and lower costs. The second is the replacement of proprietary DBMS with open source alternatives to gain vendor independence, avoid expensive licenses and audits, and gain access to the huge open source talent pool. With these two trends in mind, PostgreSQL has a tremendous opportunity for growth and a bright future.
- What is your PostgreSQL centered product?
As organizations modernize in the cloud, they’re looking to avoid the onerous lock-in associated with last generation databases, leverage their existing skills and tools and adopt open standards. An increasing number of them are standardizing on PostgreSQL as the common API for their operational databases.
To support our customers on this journey, we offer multiple choices that support the PostgreSQL client and SQL dialect, which form the key building blocks of this API.
- For easy lift & shift migrations, our Cloud SQL product provides the PostgreSQL database as an enterprise-ready fully managed service.
- For migrating or modernizing top-tier enterprise applications, AlloyDB for PostgreSQL provides a fully managed, PostgreSQL-compatible database. AlloyDB takes advantage of the cloud via separation of compute and storage, ML-enabled autopilot systems, faster transaction processing and faster analytical processing by using a columnar, in-memory accelerator.
- Finally, Cloud Spanner, our globally distributed database with unlimited scale, strong consistency, and industry leading 99.999% availability, provides a PostgreSQL interface for easy integration of existing and new applications.
- What is your company’s mission?
Google’s mission is to organize the world’s information and make it universally accessible and useful. Specifically, at Google Cloud, we want to be a trusted partner for our customers, helping them in their digital transformation journey and to solve their biggest technological challenges by building a transformation cloud.
A transformation cloud accelerates an organization’s digital transformation through application and infrastructure modernization, data democratization, people connections, and trusted transactions. The result is an organization—and its workers—that can take advantage of all of the benefits of cloud computing to drive innovation.
- What is the most annoying PostgreSQL problem, and do you have plans or ideas to fix it?
Vacuum management is a pain point we frequently hear from our customers. Even experienced database administrators struggle to find the right configuration for autovacuum. Despite spending a great deal of time with vacuum management, the fear of a database outage due to transaction ID wraparound is omnipresent.
We worked hard to address this pain point across our Postgres portfolio. Cloud SQL offers configurable alerts that notify customers early about potential vacuum related issues such as running out of transaction IDs. We also added an emergency maintenance mode which removes the need to use the single-user mode to resolve the wraparound problem.
AlloyDB further enhances and automates vacuum management. Our adaptive vacuum continuously monitors and dynamically reallocates system resources to the vacuum process without impacting database performance. If the vacuum still can’t clean up fast enough, AlloyDB throttles the rate of incoming transactions to avoid the database running out of transaction IDs and becoming unavailable.
- Which feature is missing in PostgreSQL?
Many of our customers want to combine transaction processing workloads with complex analytical queries that directly run on operational data sets and schemas. Unfortunately, the well-established Extract-Transform-Load (ETL) pattern relies on the extraction of data into a separate data warehouse, and is often complex, and error-prone. Moreover, the associated latency until data becomes visible in the warehousing system is a problem for modern, interactive applications that aim at providing real-time insights.
To enable Postgres to excel in hybrid workloads that combine transaction processing (OLTP) and analytical query workloads (OLAP), we believe Postgres needs an in-memory, columnar query processing engine.
AlloyDB for Postgres incorporates such an engine that can significantly accelerate the execution of analytical queries. It is complemented by a machine learning algorithm that continuously monitors the system workload and automatically chooses the subset of columns that benefit most from remaining in main memory.