December 2022 Summaries

Aerospike has released new hotfixes for its Database 6.2, 6.1, 6.0, and 5.7 versions, with the most recent ones being 6.2.0.2, 6.2.0.1, 6.1.0.7, 6.1.0.6, 6.1.0.5, 6.0.0.10, and 6.0.0.9. Additionally, the Aerospike REST Client has been renamed to more accurately reflect its function as an Aerospike REST Gateway. The community blog has seen recent posts on processing large data sets in parallel streams, while upcoming events are also being announced.

Data modeling with the Aerospike database allows applications to run efficiently and scale as needed. The Aerospike database is purpose-built for applications that require predictable sub-millisecond access to billions and trillions of objects and need to store many terabytes and petabytes of data, while keeping the cluster size – and therefore the operational costs – small. Data modeling in Aerospike involves additional considerations such as multi-model database, design of record keys, data modeling for related objects, understanding transactions and consistency, managing temporary objects, organizing namespaces, maximum record size, compressing data, optimizing speed, server-side operations, batch operations, secondary indexes, and more.

Data modeling with the Aerospike database is essential for efficiently mapping application objects to database models while ensuring persistence, performance, consistency, and ease of access. Aerospike is designed for applications requiring rapid access to vast amounts of data, utilizing SSDs for high-speed storage, and supporting a hybrid memory architecture. Its NoSQL nature allows for flexible schema design, emphasizing denormalization, aggregates, and single-record transactions to optimize data access and consistency. Aerospike's data modeling principles involve strategic decisions about embedding versus referencing objects based on relationship cardinality and access patterns. The database supports a multi-model approach with schema-less records, offering various data types and collection mechanisms. Key features include secondary indexes, batch operations, and server-side processing capabilities, enhancing performance and scalability. Additionally, Aerospike provides mechanisms for transaction management, namespace organization, data compression, and replication, ensuring efficient data handling and reliability across distributed architectures.

Aerospike, a Platinum sponsor at AWS re:Invent 2022, showcased its solutions and engaged with attendees through insightful presentations, hands-on demos, and productive conversations. The company's CEO led a breakout session on building real-time applications using Aerospike on AWS, featuring Fidelity Investments and Wix.com. Aerospike was also mentioned in two AWS press releases and received AWS Graviton Ready status and the AWS Data Analytics Competency Partner designation. The event proved to be exciting and rewarding for both Aerospike and its attendees, as it highlighted how the company helps businesses reduce infrastructure costs while maintaining performance at any scale.

Aerospike offers mechanisms for efficiently processing large data sets in parallel by utilizing partitioning schemes that are collectively exhaustive and mutually exclusive. These schemes allow data to be split into partitions or sub-partitions, enabling multiple worker tasks to process them concurrently. Aerospike organizes records into 4096 partitions using a hash function, ensuring uniform distribution across cluster nodes, and supports queries over these partitions with pagination for efficient data retrieval. For platforms requiring more than 4096 concurrent tasks, data can be further divided into sub-partitions using a digest-modulo function, allowing for rapid evaluation without accessing storage devices directly. The article discusses different split assignment schemes, such as At-Most N, At-Least N, and Exactly N splits, each with specific API call requirements. A parallel query framework is provided to test these assignments, accommodating various parameters like the number of splits, workers, query types, and processing modes. The potential for high degrees of parallelism is highlighted, though it notes that extreme parallelism might not always yield benefits, especially in complex computations requiring data shuffling across nodes.