November 2018 Summaries

We have made significant progress on YSQL, the PostgreSQL-compatible distributed SQL API for YugabyteDB, with updates including distributed multi-shard/multi-table ACID transactions, global & consistent secondary indexes, and native JSON data type. We also released YugabyteDB 1.1 with a ton of great features, including performance boost, public/private IP bindings for multi/hybrid cloud or multi-region deployments, user authentication, Redis-compatibility features, and support for multiple databases or namespaces. Our team has been busy writing blog posts, producing technical videos, and enhancing the documentation to provide high-quality resources for users. We've also added role-based access control (RBAC) for the YCQL API, improved serialization format of data types in YSQL, and are preparing for an upcoming release with native Kubernetes support for YSQL. Our team is growing fast, and we're looking for talented individuals to join us in Sunnyvale, CA.

The SQL vs. NoSQL debate revolves around the database core architecture and data modeling approach. Traditional SQL databases are monolithic and transactional, with ACID guarantees, but lack linear write scalability and automatic fault-tolerance. In contrast, NoSQL databases are distributed and non-transactional, with a focus on high performance and horizontal scaling, but often sacrifice ACID guarantees and consistency. To address these trade-offs, YugabyteDB brings together SQL and NoSQL on a common database core, allowing developers to model their data in SQL or queries in NoSQL, depending on the workload requirements. This approach enables complex real-world applications to utilize both SQL and NoSQL databases, promoting flexibility and agility. The choice between SQL and NoSQL ultimately depends on the specific use case, with YugabyteDB providing a viable alternative for mission-critical data.

The text discusses the performance of distributed backups in a multi-region YugabyteDB cluster, comparing it to a single region cluster. A 9-node cluster with 3 availability zones across two regions was configured for backup testing. The backup store used an Amazon S3 bucket, and the source cluster had a replication factor of 3. A read-intensive workload was loaded into the cluster before initiating the backup, which ran for about an hour without degrading performance. The test showed that distributed backups in multi-region clusters perform efficiently, with minimal impact on user experience.

YugabyteDB is a distributed database designed for high availability and performance, allowing administrators to deploy replicas in independent fault domains for improved resilience. The database supports efficient, multi-TB backups and restores under heavy read and write workloads, outperforming eventually consistent databases like Apache Cassandra. YugabyteDB's per-shard distributed consensus architecture enables lightweight checkpoint mechanisms, reducing the impact of backups on foreground operations. The database can be configured to use a cloud object storage endpoint such as AWS S3 or an NFS location for backup stores, and supports table-level backups and restores. A 6-node cluster was used as a source for backups and restores, with two scenarios tested: one during a write-heavy workload and the other during a read-heavy workload. In both cases, no major impact on foreground operations was observed, and restore operations were completed within expected times without significant performance penalties. The database's design advantages ensure highly reliable and repeatable backup/restore operations for mission-critical clusters with zero data loss.

Presto is an interactive OLAP SQL query engine that supports standard ANSI SQL queries, including complex queries and aggregations, optimized for fast execution speed. It was created by Facebook and has gained widespread adoption as a tool for interactive analytics. Presto can connect to various data sources, including SQL and NoSQL databases, Hive/Hadoop, and even join data across multiple databases easily, making it an ideal choice for deriving insights from data that has already been created by standard OLTP databases. YugabyteDB is a cloud-native, transactional, and high-performance database that supports multiple APIs, including YCQL, YEDIS, and YSQL, allowing Presto to run queries on these APIs and join data across them. By leveraging Presto on YugabyteDB, users can easily query and analyze their data in an interactive manner, making it a powerful tool for business intelligence and analytics.

MongoDB has made significant progress in addressing its durability and performance issues over the years, but it still falls short in the context of high-performance applications that require low latency and high throughput. The introduction of single-shard transactions in MongoDB 4.0 is a welcome change, but it comes with hidden costs such as lack of horizontal scalability, high latency due to the need for caching, and low throughput due to the primary node handling all write requests. In contrast, modern transactional databases like YugabyteDB deliver transactional guarantees, high performance, and linear scalability, making them a more suitable choice for fast-growing online services.