The New Stack Podcast

Platform engineering is gaining prominence due to the need for faster application deployment, which directly impacts business velocity. Valentina Alaria, Senior Director of Product at VMware, emphasizes that not all organizations pursuing platform engineering have the same goals, context, or pain points. They tailor solutions to each organization's specific needs. Some focus on rapid onboarding for junior developers, while others aim to reduce complexity, friction, and support larger development teams with fewer operational staff.Platform engineering aims to streamline collaboration between developers and operations engineers. Developers want portable code and the ability to focus on coding without worrying about production requirements. Operations engineers and platform teams seek a seamless environment for deploying applications in different contexts.Successful platform engineering initiatives involve strong collaboration models, fostering a cooperative approach rather than a siloed one. The goal is to create applications and value for the organization by facilitating effective interaction between developers and operations engineers.This podcast episode, hosted by Alex Williams of TNS, also delves into VMware Tanzu's latest tools for supporting platform engineering.Learn more from The New Stack about platform engineering and VMware Tanzu:Platform Engineering Overview, News and Trends6 Patterns for Platform Engineering SuccessA Guide to Open Source Platform EngineeringStreamline Platform Engineering with Kubernetes Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

ByConity is an open source project that emerged from ByteDance's use of Clickhouse, an open-source database system, to address their growing data volume. ByConity focuses on enhancing the separation of compute and storage, improving multitenancy support, and optimizing query performance in cloud-native environments.ByteDance's Vini Jaiswal, a principle developer advocate at the parent company of TikTok, highlights the power of open source in fostering innovation and collaboration. She shares her personal experience of leveraging open source to solve problems quickly and efficiently. She emphasizes the importance of getting involved in open source, even for those who might be hesitant, and suggests starting by identifying a pain point and making small contributions.ByConity's architecture, which separates compute and storage, offers benefits like preventing data lake corruption, read and write separation, elasticity, and scalability. Jaiswal also mentions her previous experience with open source during her time at CitiBank, where she realized how open source accelerated digital transformations.Throughout the conversation, Jaiswal underscores the strength of open source communities in collectively addressing challenges. She encourages listeners to embrace open source and start contributing, emphasizing how even small contributions can lead to significant impacts over time.The episode also delves into Jaiswal's involvement with other open source projects, such as PyTorch, and explores the intersection of open source and generative AI.Learn more from The New Stack about open source and cloud native environments:What Is 'Cloud Native' (and Why Does It Matter)?Cloud Native Ecosystem News and ResourcesHow to Build an Open Source Community Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Along with discussing the emergence and ascension of platform engineering in this episode, we also discuss the role that Humanitec plays in helping organizations establish platforms for developers, as well as Backstage, a popular open source internal developer platform that was developed by Spotify for its own developers.An IDP, our guest Kaspar Von Grünberg explained, is a standardized interface for developers to build applications using a  golden path of vetted tools and libraries, allowing for a high degree of efficiency for both the developers themselves as well as the engineers who are supporting the developers. They can include an integration and delivery plane, a continuous integration registry, a platform orchestrator, observability tools and a resource plane."How you're consuming this is a little bit up to the individual preference of the user, and what the platform team has configured for you. So we're seeing some teams like to use a user interface and some teams like to use code based interactions," Von Grünberg explained.In some ways, a IDP is reminiscent of the platform-as-a-service packages of a decade ago. They also were designed to help developer efficiency, though devs chafed at the limited number of tools they were allowed to use  in these walled gardens. That was a mistake, Von Grünberg said.Those platforms required developers to use a small set of pre-defined times."We don't want to get back to those times, which is why we want to provide sensible defaults," Von Grünberg said. A good IDP will provide developers with "golden paths" or "paved roads" as Netflix calls them."Developers can stay on those paths if they want," Von Grünberg said. They can enjoy the security default and service-level agreements (SLAs) from the engineers. But developers are also free to leave the path and make low-level configurations on their own as well."Good platform engineering is never about covering all the use cases," he said.Learn more from The New Stack about platform engineering and Humanitec:Platform Engineering Overview, News, and TrendsHow to Pave Golden Paths That Actually Go SomewhereBuild Your IDP at Light Speed with a Platform Reference Architecture Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

John Willis, a technologist and author, cautions against blindly following vendor recommendations for AI products. He suggests a "DevOps redo" to encourage experimentation and collaboration. Willis highlights the importance of educating teams on data management techniques, including retrieval augmentation. He emphasizes the need for data cleansing to prevent undesirable code or sensitive information. The podcast also discusses the rise of Shadow AI, the difference between co-completion tools and proprietary tools, and the challenges of responsible AI implementation.

Deliveroo, a prominent food delivery company, relies on Apache Flink, a distributed processing engine, to enhance its three-sided marketplace, connecting delivery drivers, restaurants, and customers. Seeking to improve real-time data streaming and gain insights into customer behavior, Deliveroo transitioned to Flink, comparing it to alternatives like Apache Spark and Kafka Streams. Flink, with feature parity to their previous platform, offered stability and scalability. They initially experimented with Flink on Kubernetes but turned to the Amazon Managed Service for Flink (MSF) for enhanced support and maintenance.Engineers from Deliveroo, Felix Angell and Duc Anh Khu, emphasized the need for flexibility in data modeling to accommodate their fast-paced product development. However, flexibility can be complex, often requiring data model adjustments. They expressed the desire for a self-serve configuration feature in MSF, allowing easy customization of low-level settings and auto-scaling based on application metrics. This move to Flink and MSF has empowered Deliveroo to focus on core responsibilities like continuous integration and delivery while efficiently managing their data processing needs.Learn more from The New Stack about Apache Flink and AWS:Kinesis, Kafka and Amazon Managed Service for Apache FlinkApache Flink for Real Time Data AnalysisApache Flink for Unbounded Data Streams Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Over the past five to ten years, the testing of microservices has seen significant growth. This surge in testing can be attributed to the increasing adoption of microservices and Kubernetes, which signify a shift away from monolithic application architectures. Bruno Lopes, a leader at Kubernetes company incubator Kubeshop, noted this trend. Kubeshop has initiated six Kubernetes projects, including TestKube, a Kubernetes native testing framework led by Lopes.This rise in testing is making it more accessible to a wider audience and is enhancing the developer experience through automation. Developers now have more time to focus on innovation rather than manual testing. However, there is often a disconnect between development and testing, as developers move quickly, outpacing organizational adaptation to modern testing methods.Lopes emphasized the importance of testing before production deployment and advocated for creating production-resembling testing environments that allow for rapid deployment without waiting for manual tests. This approach is particularly critical for Site Reliability Engineering (SRE) teams who need to respond quickly to issues and minimize downtime for customers. In some cases, it's necessary to run tests within Kubernetes itself, a concept that may take time for companies to fully embrace as the developer experience continues to improve.Learn more from The New Stack about Kubernetes, Testing and TestKube:Testkube: A Cloud Native Testing Framework for KubernetesTop 5 Challenges in Modern Kubernetes TestingWhy You Should Start Testing in the Cloud Native Way Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Deepthi Mohan and Nagesh Honnalii from AWS discuss Apache Flink and Amazon Managed Service for Apache Flink. They highlight use cases like streaming ETL, real-time analytics, and complex event processing. They trace the origins of MSF back to Amazon Kinesis and Apache Kafka. They also discuss the advantages of using a managed service for Flink, such as ease of use, scalability, and security. The podcast also covers how to get started with Amazon Managed Service for Apache Flink, including pricing details.

Modern developer conferences like the upcoming Infobip Shift Conference in Croatia are centered around themes. At this particular event for developers, you can expect a lot of focus to be on the developer experience and artificial intelligence (AI).Ivan Burazin, Chief Development Experience Officer at InfoBip, joined us on the show and emphasizes that developers spend a substantial portion of their time not coding, often losing 50 to 70% of their productive hours to non-coding activities, such as setting up environments, running tests, and building code. This highlights the importance of improving the developer experience to enhance productivity.The developer experience has both internal and external dimensions. Externally, it impacts customer experience, while internally, it influences development velocity. A better developer experience translates to faster and more efficient coding.The Shift Conference will feature talks on six stages, one of which will focus on the developer experience, addressing its internal and external aspects. Additionally, AI will take center stage at another segment of the conference.Although there may not be an abundance of true AI experts taking the stage, the focus will be on how individuals and companies can leverage AI to create products and services. It's recognized that AI will play a pivotal role in the future of every industry, and the conference aims to explore practical applications and strategies for integrating AI into various businesses.Overall, the Shift Conference aims to address the challenges developers face in optimizing their productivity and explore the growing importance of AI in shaping the future of businesses and products.Learn more from The New Stack about the developer experience and InfoBip Shift:7 Principles and 10 Tactics to Make You a 10x DeveloperThe Challenges of Marketing Software Tools to DevelopersA Guide to Better Developer Experience Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

This episode delves into Apache Flink, a versatile platform for executing both batch and real-time streaming data analysis tasks. This session marks the beginning of a three-part series unveiling Amazon Web Services' (AWS) new managed service built on Flink. Future episodes will explore this service in detail and examine customer experiences.The podcast features insights from Danny Cranmer, a principal engineer at AWS and an Apache Flink PMC and Committer, along with Hong Teoh, a software development engineer at AWS.Flink stands out as a high-level framework for defining data analytics jobs, accommodating both batch and streaming data sets. It offers APIs for building analysis jobs in various languages, including Java, Python, and SQL. Flink also provides a distributed job execution engine with fault tolerance and horizontal scaling capabilities.One prominent use case is Extract-Transform-Load (ETL), where raw data is swiftly processed for specific workloads. Flink excels in delivering low-latency transformations for unbounded data streams. Additionally, Flink supports event-driven applications, responding immediately to triggers such as user requests for weather data.Flink ensures exactly-once processing, critical for scenarios like financial transactions. It employs checkpoints to maintain data integrity in case of node failures.The podcast also touches on AWS's role in supporting the open-source Flink project and the future outlook for this powerful data processing framework.Learn more from The New Stack about Apache Flink:3 Reasons Why You Need Apache Flink for Stream ProcessingApache Flink for Unbounded Data Streams8 Real-Time Data Best Practices Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Renowned technologist Adrian Cockcroft discusses the process of fine-tuning Large Language Models (LLMs) through prompt engineering. Crafting specific prompts guides AI's output. Fine-tuning of LLMs using internal data like wiki pages enables better domain and process understanding. Vector databases are in demand for enhanced information retrieval from LLMs. Cockcroft explores the challenges and potential pitfalls of code-generating assistants. He also highlights the importance of embracing new technologies and startups in the fast-moving tech field.