April 2018 Summaries

In the ninth installment of the Mutant Monitoring Systems blog series, the focus is on teaching Division 3 developers how to create applications that interact programmatically with a ScyllaDB cluster, utilizing its compatibility with Apache Cassandra to leverage existing libraries in languages like Node.js and Go. The post guides developers through setting up a sample Node.js application to connect with the ScyllaDB cluster, highlighting the importance of database connectivity in application development. By using the cassandra-driver library, developers can perform operations such as INSERT, DELETE, and SELECT within ScyllaDB's keyspace, ensuring resilience by connecting to multiple nodes and enhancing performance through load balancing. The tutorial also covers automating data insertion using a setTimeout loop and demonstrates the process of setting up and running the ScyllaDB cluster within a Docker environment, concluding with instructions on verifying the successful execution of data operations within the system.

ScyllaDB, led by CEO Dor Laor, has secured a new round of funding, bringing its total to $35 million, enabling the company to expand its sales and marketing efforts and accelerate the development of ScyllaDB as a service. The company has witnessed significant growth, with a diverse clientele ranging from Fortune 50 companies to startups, benefiting from ScyllaDB's high performance, real-time Big Data processing capabilities, and its ability to scale efficiently, often reducing server costs and maintenance time compared to other databases like Apache Cassandra. ScyllaDB continues to innovate and enhance its offerings, focusing on performance, manageability, and automation, while maintaining compatibility with Cassandra and advancing its Database-as-a-Service (DBaaS) initiative to meet global demand. The company also emphasizes its commitment to providing a flexible, open-source platform that avoids vendor lock-in and supports multi-cloud deployments, ensuring improved reliability and cost savings for its users.

ScyllaDB Manager 1.1, a production-ready release for ScyllaDB Enterprise customers, enhances cluster administration with centralized management and automation of tasks like periodic repairs. It introduces new features such as metric reporting through the Prometheus protocol, a dashboard for monitoring, and alerts for repair failures or manager exits. The update includes configurable repair retries, improved date formats, and REST API enhancements. Additionally, the release addresses bug fixes, such as clarifying error messages and allowing task lists to display across all clusters. ScyllaDB Manager is available for download with a 30-day trial, and it is set to become a central tool for cluster management with plans for a GUI frontend in future updates.

ScyllaDB 2.1.2, released by the ScyllaDB team, is a bugfix update for the ScyllaDB 2.1 stable branch, ensuring backward compatibility and support for rolling upgrades. This update addresses several issues, including problems with view entries having non-composite, empty partition keys, and node removal failures during the bootstrap process. It also resolves a TLS session closure error, a NullPointerException with nodetool gossipinfo during cluster initialization, and limitations in stableloader and sstableloader concerning user-defined types and binding prepared statements for tables using the "date" type. Users are encouraged to report any problems encountered with the release, available through Docker, binary packages, and EC2 AMI.

ScyllaDB's I/O Scheduler plays a crucial role in managing workloads and ensuring efficient operations by prioritizing requests to modern SSDs, optimizing latency, and maintaining throughput. The scheduler identifies the maximum useful disk concurrency to prevent latency spikes caused by queuing excess requests at the disk level, instead managing them within the database. Recent improvements address issues such as the differentiation between read and write speeds, request size considerations, and better user understanding of throughput. By applying Little’s Law, ScyllaDB can dynamically adjust concurrency to maintain optimal performance, reducing latency significantly in various scenarios. These enhancements are set to be included in ScyllaDB 2.3, promising even better performance for latency-sensitive workloads.

The blog post, part of a series on ScyllaDB training, focuses on setting up and using ScyllaDB Monitoring to oversee a Mutant Monitoring System (MMS) cluster, following an incident where a datacenter was destroyed due to a lack of proper monitoring. It details the process of configuring the ScyllaDB Monitoring stack, which employs Prometheus for gathering metrics and Grafana for visualizing data, all running in Docker containers on a shared virtual network. The post explains how to start the ScyllaDB cluster, configure the monitoring tools, and utilize a load generator to simulate activity within the system, thereby producing more informative graphs. It also describes how to handle node availability issues, demonstrating the dashboard's capability to quickly identify and resolve such problems, ensuring effective monitoring of the MMS infrastructure.

Samsung SDS, a global IT services company, faced performance challenges with their relational databases, prompting them to explore NoSQL alternatives for superior throughput, scalability, and reduced costs. During a technical evaluation, they compared ScyllaDB with Apache Cassandra, discovering that ScyllaDB provided three times better throughput and latency. The evaluation also highlighted ScyllaDB's advantages in cost savings and maintenance ease, leading Samsung SDS to adopt it for enhancing their database performance. Principal Data Architect Kuyul Noh emphasized the cost-effectiveness of ScyllaDB over other NoSQL options, underscoring its excellent performance in meeting Samsung's business needs.

IBM's Open Tech and Performance teams conducted a series of tests to evaluate ScyllaDB against Apache Cassandra and HBase in terms of performance and scalability for use in IBM Cloud's Compose for JanusGraph. The tests revealed that ScyllaDB significantly outperformed its counterparts, achieving nearly 35% higher throughput than HBase and almost three times the throughput of Cassandra in the first test, while in the second test, it showed 160% better throughput than HBase and more than four times that of Cassandra. ScyllaDB also proved to be the easiest to cluster, with user-friendly features like self-tuning capabilities and efficient load balancing that allow it to maximize system resource utilization. The IBM Compose team expressed satisfaction with ScyllaDB's performance advantages and clustering ease, encouraging further exploration of its capabilities.

ScyllaDB's upcoming open-source release will incorporate role-based access control (RBAC) to manage authentication and authorization, enhancing security beyond the default unrestricted access. Traditionally, user-based access control could become cumbersome, especially in settings with numerous resources requiring similar permissions, potentially leading to non-repudiation issues if an "umbrella" user is created. RBAC introduces roles as a solution, allowing permissions to be assigned to roles, which can then be granted to users, streamlining permission management and maintaining traceability. The transition to this system involves the automatic migration of existing user-based metadata to a new schema during a cluster upgrade, ensuring continuity and improved security without manual intervention. This enhancement aligns ScyllaDB with Apache Cassandra's approach while maintaining compatibility with older user-based commands for ease of migration.

IMVU, a popular social platform with over 100,000 concurrent users, transitioned from a database architecture utilizing Memcached, MySQL, and Redis to ScyllaDB to meet its growing scalability needs. The company found ScyllaDB easy to install, maintain, and scale, achieving excellent performance results quickly. Ken Rudy, a senior software engineer at IMVU, praised the responsive support from ScyllaDB, noting quick assistance even before finalizing their enterprise contract. This transition was motivated by the need for a more efficient and scalable database solution to support IMVU's global user interactions through 3D avatars across various devices.

In a recent webinar featuring Brian Hawkins, the creator of KairosDB, various questions about building a time-series database using KairosDB and ScyllaDB were addressed. It is possible to use a single node of each for development, but for production, multiple nodes of both are recommended to ensure high availability and resiliency. KairosDB requires Java 1.8 or higher and certain server specifications, while ScyllaDB suggests a minimum of 4 CPUs, 8GB RAM, and 1TB SSD for optimal performance. The combined solution of KairosDB and ScyllaDB offers advantages over ElasticSearch, including easier administration, maintenance, and better throughput and latency, with the added capability of setting TTL for data. The webinar also discussed the ability to configure KairosDB for high availability and scalability, the lack of special time-series support in ScyllaDB but its efficient handling of wide rows, and potential latencies when reading data across nodes. While a direct pass to backend storage is being developed for KairosDB, the current setup effectively queues data, and discussions continue on the KairosDB mailing list for further updates.