December 2019 Summaries

Ethan J. Jackson's blog post explores how microservices and continuous integration and delivery (CI/CD) have enabled the development of Kubernetes, which facilitates multi-cloud environments by offering a standardized, portable interface, thus overcoming traditional vendor lock-in imposed by cloud providers. While this standardization allows for easier application migration across different cloud providers, it introduces significant challenges for developers due to increased complexity, as they must navigate numerous dependencies and varied tools for local testing versus production. Despite these difficulties, Jackson argues that these challenges can be mitigated by creating custom development environments that are fast, efficient, approachable, and closely resemble production settings, thereby enhancing developer productivity while retaining the benefits of microservices. Kelda, a development environment being built by Jackson's company, aims to address these issues by providing a platform that aligns closely with production environments, making it easier for developers to test and deploy their code efficiently.

Education First (EF), a global education company with a presence in over 50 countries, partnered with Cockroach Labs to enhance their digital learning platform by using CockroachDB Dedicated, a fully managed relational database. As EF expanded, their existing database infrastructure, which included MySQL and Cassandra, struggled to meet their global needs, prompting a search for a more scalable and efficient solution. CockroachDB Dedicated was chosen over Amazon Aurora due to its ability to handle synchronous global transactions, offering EF developers a seamless and efficient database experience without the burden of scalability issues or operational complexity. The transition allowed EF's development team to focus on building applications like the new Teachers First platform, leveraging CockroachDB’s PostgreSQL wire compatibility to simplify the process. This collaboration is detailed in a case study that highlights how EF overcame technical challenges to improve developer efficiency and support their global educational mission.

Cockroach Labs' 2020 Cloud Report provides an in-depth analysis of cloud performance, comparing AWS, Google Cloud Platform (GCP), and Azure through over 1,000 benchmark tests, including CPU, network throughput, storage I/O, and TPC-C, a tool simulating e-commerce workloads. The findings reveal that GCP has notably improved its performance, bringing it closer to AWS and Azure, particularly in the TPC-C benchmarks where all three clouds now exhibit similar top-end performance. The report also emphasizes the significance of efficiency, with GCP offering the best performance per dollar among tested machine types, though AWS excels purely in throughput. Additionally, the report introduces Azure to the benchmarking process and shares open-source tools like Roachprod for replicating tests, underlining the commitment to transparency and reproducibility in cloud performance evaluation.

In a blog post by Dan Kelly, the complexities and challenges of implementing chaos tests in multi-cloud environments are explored, particularly in the context of e-commerce systems stressed during peak shopping events like Black Friday and Cyber Monday. Ana Medina, a chaos engineer from Gremlin, highlights the importance of proactively exposing bugs through chaos testing by assuming cloud vendor failures and testing basic functionality, data consistency, and cost implications. The discussion emphasizes that while multi-cloud infrastructure can mitigate outages, it introduces new challenges that require thorough chaos testing to ensure a robust failover strategy. Talks from ESCAPE/19, including those by Dan Papandrea and Spencer Kimball, further elaborate on securing services and handling data across global multi-cloud deployments, underlining the necessity of chaos engineering in verifying multi-cloud strategies.

Peewee ORM, a lightweight and intuitive Python Object-Relational Mapping tool, has been updated to support CockroachDB, a distributed and horizontally-scalable SQL database. This new compatibility allows developers to easily configure a CockroachDatabase instance and leverage its SQL API, which is similar to Postgres, including features like jsonb and arrays. Peewee's integration with CockroachDB also supports client-side transaction retries through a helper method, enabling seamless execution of transactional SQL statements with automatic retries until a transaction is successfully committed. The update aims to simplify the use of distributed databases, making it more accessible for developers to scale their applications with minimal effort.

The blog post from Cockroach Labs explores the benefits of using CockroachDB's Follower reads feature to reduce latency in multi-region applications, using a demo application called Wikifeedia as an example. Wikifeedia, built on Wikipedia's public APIs, is designed to provide users worldwide with a low-latency experience by accepting slightly stale data, making it suitable for Follower reads. The feature allows replicas to serve read requests from the nearest location, reducing latency by avoiding the need to communicate with a potentially distant leaseholder. Initially, Wikifeedia's deployment faced high latencies due to centralized database configurations; however, by implementing a global deployment with Follower reads, significant improvements were achieved, notably reducing latency for users in regions like Singapore. The post emphasizes the importance of global application and database deployment, alongside caching and load balancing strategies, for optimizing performance in global applications, and highlights Follower reads as one of many tools available in CockroachDB's multi-region capabilities.