March 2015 Summaries

This week marked the successful Couchbase Live Europe event in London, featuring discussions on SQL for Documents (N1QL) and multi-dimensional scaling strategy. The company is now preparing for its main global user conference, Couchbase Connect 2015, to be held at Levi's Stadium in Santa Clara from June 2-4. Key areas of focus for the event include integrating Couchbase with other technologies, practical application development experiences, devops strategies, and mobile and IoT applications. The deadline for talk submissions is April 17th, with selected speakers to be notified in early May.

Multi-Dimensional Scaling (MDS) is a distributed database scaling approach that separates and scales individual services such as query, index, and data to improve performance and resource utilization. This allows for distribute data without distributing query execution and indexes, optimizing hardware usage by running different services on different nodes with varying requirements. MDS eliminates resource contention by isolating services to prevent queries and indexes from slowing down reads and writes. It improves performance by storing data and indexes on separate nodes, executing queries on separate nodes, and reducing network overhead. Additionally, it optimizes resource utilization by configuring query service nodes with fast processors, index service nodes with fast SSDs, and data service nodes with more memory. MDS offers flexible deployment options, allowing administrators to decide whether or not to leverage the technology. The approach addresses limitations of traditional sharding methods, such as custom-column sharding, which can lead to cross-node traffic for basic queries. Instead, MDS enables automatic-range-partitioning, reducing network overhead and improving scalability.

MongoDB 3.0, featuring the optional WiredTiger storage engine, was benchmarked against Couchbase Server by Avalon Consulting to assess performance improvements. While MongoDB claims WiredTiger offers significantly faster write performance than the default MMAP storage engine, the benchmark revealed that Couchbase Server outperformed MongoDB, particularly in terms of throughput, with Couchbase maintaining lower latency under higher client loads. The benchmark, conducted on Amazon Web Services using the Yahoo! Cloud Serving Benchmark, demonstrated that Couchbase Server could handle more concurrent clients while keeping latency under 5ms, a threshold MongoDB exceeded with fewer clients. The test highlighted that MongoDB's limitations were more related to sharding than the storage engine itself, suggesting that MongoDB would require significantly more nodes to match Couchbase's performance, pointing to concurrency constraints as a limiting factor in MongoDB's architecture.

The goal of this series of blog posts is to help readers understand how to set up Couchbase Server on Windows Azure, starting with a single node developer setup and progressing to a production-ready setup with multiple nodes and failover. To follow along, readers need an Azure subscription, which can be obtained for free. The first post sets up a single node Couchbase Server for testing and development purposes, including creating a Virtual Machine, installing Couchbase Server, configuring the VM's security settings, and setting up the Couchbase Console. The post also mentions that the recommended hardware for Couchbase is 4 cores and sufficient RAM, but notes that a smaller VM like A1 can be used for setup and configuration purposes. The author has run tests on their own single node Couchbase Server and obtained results of around 75-125 ops per minute with full writes only, noting that network latency could be minimized by running the test from an Application Server on the same Virtual Network as the Cluster.

Couchbase is preparing to release the .NET SDK version 2.1.0, which will introduce significant enhancements such as support for async/await keywords and a new non-blocking IO model, promising up to 10 times the performance improvement over the synchronous API. The anticipated release is set for the first week of April 2015. Meanwhile, version 2.0.3 has been released as a maintenance update, addressing various bugs and improvements, including the addition of async methods for ClusterManager and BucketManager classes, and configuring operation lifespan. Notable changes include the renaming of the ConnectionTimeout property to SendTimeout and alignment of the .NET Prepared Statement API with the Java SDK. The packages for version 2.0.3 are available on NuGet, S3, or directly from the master source.

The first Couchbase Hangout on Air of 2015 showcased the live broadcast capabilities of Google Hangouts, which allows producing live TV-like shows and automatically creating a YouTube video afterward. During this session, participants explored running Couchbase on Azure, with the main highlight being the introduction of the first developer preview of the Couchbase Spark connector by Michael Nitschinger. Viewers were encouraged to watch the hangout on YouTube and provide feedback or suggestions for future sessions through comments or forums, while updates on upcoming hangouts were to be shared via Twitter and Google Plus.

Beginning in March, Couchbase Developer Advocates will be broadcasting live, offering insights into their team's plans, past activities, and recent developments in Couchbase. The broadcasts aim to engage with the community by including demonstrations, user testimonials on innovative uses of Couchbase, and Q&A sessions with Couchbase engineers. Audience participation is encouraged, with an open invitation for viewers to share their ideas on topics they would like covered.

The Spring Data Couchbase framework allows developers to simplify interactions with Couchbase using standard Spring Data features, including caching, document validation, and exposing a repository through a REST API. To implement caching, developers can use the `@Cacheable` annotation to specify a cache instance and key regex, ensuring that method results are fetched from the cache instead of re-executing them. Document validation is achieved by adding Hibernate validation annotations to POJO objects, which throw ConstraintViolationExceptions when invalid data is attempted to be stored. The Spring Data REST Web MVC dependency enables developers to expose their repository behind a REST API, providing HATEOAS-style links and embedded documents in the response.