September 2016 Summaries

Presto is an open-source, distributed ANSI SQL query engine designed for running ad-hoc and interactive analytic queries on a variety of data sources, including ScyllaDB. Unlike Apache Hive, it does not rely on Hadoop's MapReduce, allowing it to efficiently execute SQL queries across multiple databases, which is particularly beneficial for organizations with diverse data systems. Presto's simple installation and compatibility with ScyllaDB via a shared connector with Cassandra highlight its ease of use and enhanced performance capabilities. At the ScyllaDB Summit 2016, a demonstration showcased how Presto and ScyllaDB can be integrated using Docker for analytics on air quality data, emphasizing Presto's capability to handle large data sizes from gigabytes to petabytes with impressive throughput and latency. The event encouraged further discussion and engagement through the ScyllaDB user group and social media channels.

ScyllaDB 1.3.1, released by the ScyllaDB team, is a bugfix update to the ScyllaDB 1.3 stable branch, maintaining backward compatibility and supporting rolling upgrades. Key issues addressed in this release include fixing crashes caused by reading large columns or from slow disks, resolving network setup problems affecting performance on servers with fewer than eight v-cores, and correcting UUID generation problems that impacted compatibility with certain drivers. Other fixes include addressing service startup order issues on server boot, resolving Thrift query discrepancies that affected Presto and Spark users, and improving task estimation accuracy in compaction management. The release also introduces several enhancements and cleanup operations, such as improving Docker image descriptions, updating submodules, and refining configuration scripts for better network and service handling.

ScyllaDB 1.2.5, a bugfix release of the ScyllaDB 1.2 stable branch, has been announced by the ScyllaDB team, bringing backward compatibility and support for rolling upgrades. This release addresses various issues, including the inability of scylla.yaml to read initial_token strings with spaces and unexpected aborts when querying with contradicting clustering column restrictions, which are now fixed to return an empty result. Contributions from several developers, such as Paweł Dziepak and Tomasz Grabiec, have led to improvements like ignoring spaces in the initial_token list and fixing use-after-free issues involving bound_views. The release also provides guidance for upgrading from 1.1.x to 1.2.y and encourages reporting problems and contributing to ScyllaDB development.

ScyllaDB is a database designed to maximize resource utilization on a machine, making it highly efficient for production environments by consuming all available disk, network bandwidth, RAM, and CPU resources. However, in development and testing environments where resources are shared, adjustments are necessary to prevent performance issues on other processes or user interfaces. The article provides guidance on configuring ScyllaDB to use less memory and CPU by setting specific parameters in configuration files or command line options, such as limiting memory usage and logical cores, and enabling an overprovisioned mode. Additionally, it suggests using --developer-mode to bypass unnecessary hardware and OS checks in non-production settings, and mentions the potential use of Docker for ScyllaDB deployments in development scenarios, hinting at a future discussion on this setup. Overall, while ScyllaDB's default setup is optimized for performance, it can be easily adapted for shared environments to ensure smooth operation alongside other applications.