April 2021 Summaries

This webinar presented how to run distributed SQL databases on Kubernetes, focusing on the design of stateful workloads in Kubernetes, the architecture and deployment of YugabyteDB in Kubernetes, and best practices for running cloud-native stateful workloads. It highlighted the adoption of Kubernetes in enterprise Fortune 500 companies, containerized data workloads offering advantages such as better cluster resource utilization, portability between cloud and on-premises, self-service experience, and robust automation frameworks. The presentation covered what is a distributed SQL database, its resilience to failures, horizontal scalability, and geographical distribution. YugabyteDB was deployed in Kubernetes using StatefulSets, with architectures for synchronous and asynchronous replication topologies. The webinar also showcased the deployment of Yugabyte Platform, a self-service portal that enables developers to spin up a database for their apps in minutes, simplify day-2 operations through automation, and get services needed to realize business outcomes with the database. Key highlights included resilience to node failures, scaling out, and cloud-native microservices application demos.

The YugabyteDB community delivered a free "YSQL Development Intermediate" training and certification class on April 22, 2021. The 90-minute course covered various topics such as YugabyteDB review and setup, constraints, UDFs, expressions, and operators, indexes, views, triggers, extensions, and more. Over 40 exercises were worked through to demonstrate how developers can apply their PostgreSQL knowledge to YugabyteDB. The certification exam was moderately challenging, with an average rating of 3 out of 5, indicating that careful studying is required to pass. The community also addressed several questions from attendees, including migration from IBM DB2, fine-tuning the default sharding strategy, and recovering a dropped database. The next training and certification in the series is scheduled for May 27, with registration available for free.

You can use the COPY command with the ROWS_PER_TRANSACTION attribute to import large tables in smaller batches, reducing memory pressure and avoiding errors. YugabyteDB offers various migration options for zero-downtime migrations, including using Blitzz or Debezium, and utilizing tools like pg_dump/pg_restore and pgloader. Changing default ports can be done by adding flags to the configuration file, while Helm chart values can be modified to configure the cluster according to specific requirements. New documentation, blogs, tutorials, and videos have been added, including resources on creating a centralized consent database, getting started with DBeaver EE and Yugabyte Cloud Query Language, and running distributed SQL databases on Kubernetes. Upcoming events include certification training, DBA fundamentals training, and conferences like KubeCon + CloudNativeCon Europe 2021 Virtual. The documentation is continually updated to provide the necessary information for making the most out of YugabyteDB.

Softtech, a 1,500+ person technology company based in Turkey, presented a use case for creating a centralized consent database to manage over 80M citizens' preferences across 100k companies doing business in Turkey. The database required strong consistency, horizontal scalability, and support for high throughput and low read latency. Softtech explored MongoDB, Cassandra, and YugabyteDB as possible solutions and ultimately chose YugabyteDB for its scalability, microservices-based architecture, and event-driven design. With YugabyteDB, the company achieved significant technical and business outcomes, including managing 500 million transactions per day, 10 million transactions per minute at peak times, and storing over 5 billion consents. The database is designed to be available, encrypt personal information, and support multi-region replication for a subset of data.

YugabyteDB has made significant changes to its YSQL API, including the removal of client certs for authentication in TLS enabled clusters. The default configuration now uses password authentication instead. Three main flags related to TLS are `ysql_enable_auth`, `use_client_to_server_encryption`, and `ysql_hba_conf_csv`. These flags can be used to manually configure a custom HBA (Host-Based Access Control) configuration, which can help with security and performance. Additionally, YugabyteDB has made changes to its role inheritance mechanism, which can cause issues when creating new roles. The YSQL API also supports crosstab queries, also known as pivot queries in Excel, using the tablefunc extension. Furthermore, YugabyteDB offers a choice between having multiple databases with one schema each or one database with multiple schemas, each with their strengths and weaknesses. Finally, the documentation for YugabyteDB is continually being updated and improved, including new content added and changes made to existing documentation.

CAST AI is an AI-driven cloud optimization platform for Kubernetes that helps companies cut their cloud bill and prevent downtime. At this year's Distributed SQL Summit Asia 2021, Leon Kuperman from CAST AI presented a talk titled "An Adventure in Multi-Cloud with Kubernetes and Distributed Data." The presentation demonstrated how to deploy a highly available YugabyteDB cluster across three cloud providers (AWS, GCP, & DigitalOcean) using Kubernetes and CAST AI automation. The architecture consisted of multiple exposed load balancers and ingress controllers acting as the entryway into each of the three clouds. Inside each cloud environment, there was a highly available Kubernetes cluster configured with CAST AI, which was wired to YugabyteDB as the distributed backend. The presentation also covered the deployment of an e-commerce application using Kubernetes, highlighting the importance of persistence, latency, and performance in a multi-cloud setup. Leon demonstrated how to deploy the application, configure DNS and routing correctly, and showcase the benefits of using CAST AI and YugabyteDB for a highly available and performant e-commerce application. The presentation also included a TPC-C benchmark, which showed that the multi-cloud scenario performed well compared to the mono-cloud scenario, with slightly higher latency due to cross-cloud communication. Overall, the talk highlighted the potential of combining Kubernetes, distributed data, and CAST AI to create highly scalable and available cloud deployments.