October 2016 Summaries

ScyllaDB version 1.3.3, released by the ScyllaDB team in October 2016, is a bugfix update to the 1.3 stable branch, maintaining backward compatibility and supporting rolling upgrades. This release addresses several issues, including problems with the CQL GRANT statement, server crashes when the cache is disabled, RPC connection drops on error responses, and broken clustering caused by the sstableloader. Contributions from developers such as Glauber Costa, Pekka Enberg, and Tomasz Grabiec focused on improving authentication processes and fixing partition version list corruption. The update emphasizes enhancements for a more stable and efficient ScyllaDB experience.

Compose's Database as a Service (DBaaS) platform simplifies infrastructure management, allowing operators and developers to focus on application innovation by offering ScyllaDB, a highly efficient NoSQL solution known for its speed and availability. ScyllaDB provides significant latency reduction for developers and lessens the management burden on operators by reducing necessary components, making it an attractive option for applications demanding fast and reliable databases. A webinar titled "ScyllaDB on Compose" scheduled for November 9th aims to provide an overview of ScyllaDB, highlighting its efficient use of hardware, minimal maintenance due to its C++14 architecture, and compatibility with both CQL and Thrift protocols, which ensures seamless workload transition to a Compose-managed service without requiring application modifications.

Dor Laor and Glauber Costa will present a tech talk at AppNexus in New York on November 25th, focusing on ScyllaDB's workload conditioning techniques, which ensure stable performance through automatic performance tuning. ScyllaDB is a high-performance alternative to Apache Cassandra, boasting wire compatibility and being ten times faster, achieved through advanced features like its own caching, malloc, log structured allocator, and a user-space TCP/IP stack. Despite these capabilities, ScyllaDB's distinguishing factor is its ability to handle various challenges such as failures, spikes, timeouts, and out-of-memory issues by dynamically adjusting system parameters in response to workload fluctuations. The talk will delve into these low-level aspects and techniques, underscoring that speed alone is insufficient without effective control, and attendees are invited to explore these insights further at the ScyllaDB Summit 2016.

ScyllaDB 1.3.2 is a bugfix release for the ScyllaDB 1.3 stable branch, maintaining backward compatibility and supporting rolling upgrades. This update addresses several issues, including missing information in nodetool cfstats reports when using LeveledCompactionStrategy, incorrect handling of clustering row ranges in paging code, inconsistent gossip expiration timers, and request throttling to prevent memory overflow. Additional fixes ensure proper handling of gossip failure_detector history and prevent premature timeouts for sstable readers during streaming. Contributions from various developers involved changes to the gossip system, query pagers, and database operations to enhance the reliability and efficiency of data handling and processing.

ScyllaDB's Slow Query Logging is a valuable tool for identifying and addressing performance issues related to slow query responses by capturing execution details, allowing for a deeper understanding of query bottlenecks. The feature enables users to trace not only the problematic query but also its intermediate timing traces, offering insights into why a query might be slow. Configuring the Slow Query Logging involves setting parameters like time-to-live (TTL) and a threshold for query duration, all managed via a REST API. The article demonstrates a practical scenario with a cluster set up on AWS, highlighting how changing the consistency level can resolve latency issues, and discusses the performance impact of enabling this feature, noting that it incurs a CPU usage penalty, especially under high load conditions. Despite its potential impact on throughput, the tool is useful for environments where the system's load is manageable, providing essential visibility into query execution without significantly affecting latency.

In a talk at ScyllaDB Summit 2016, key insights into effectively monitoring a ScyllaDB cluster were shared, emphasizing the importance of understanding performance characteristics, overall health, and potential issues. ScyllaDB, operating as a regular Linux process, can produce misleading outputs when monitored with standard Linux tools like 'top' due to its busy-looping technique aimed at minimizing latency. To counteract this, ScyllaDB exports internal metrics that can be integrated with tools such as collectd, or its own tools like scyllatop for immediate metric checking. Additionally, ScyllaDB offers a Prometheus and Grafana solution with pre-loaded dashboards for comprehensive system monitoring. Understanding these metrics and their implications is crucial for successful ScyllaDB deployment, and resources such as videos and slides from the presentation provide further guidance on this topic.