Cloud Engineering Archives - Software Engineering Daily

The intent of the DevOps movement is to get organizations moving faster and more effectively by breaking down siloes, and improving communication. Gene Kim’s book The Phoenix Project illustrated this by telling the fictional story of a company adopting a DevOps mentality. Although that book was fiction, Gene is an experienced engineer, having worked as founder and CTO of Tripwire, a software company that makes security and compliance automation software. In his new book The DevOps Handbook, Gene presents a practical companion to The Phoenix Project. Together with his co-authors, Gene has written a guide for how to move an organization toward DevOps, and in this episode we explore some of the topics from his book. Check out an excerpt from Gene’s book here: DevOps Handbook The post DevOps Handbook with Gene Kim appeared first on Software Engineering Daily.

  At Netflix, developers write applications with a variety of requirements–from simple requests for a list of movies to more resource-intensive requests like a complex machine learning workflow. Netflix wants developers to be able to request the resources they need from a compute cluster and receive those resources on-demand, without thinking too much about the state of that pool of resources they are drawing from. At the cluster level, this means scheduling the application requests intelligently. Sharma Podila, a distributed systems software architect at Netflix, explains how Netflix has built a cloud native scheduling system on top of Mesos. The post Netflix Scheduling with Sharma Podila appeared first on Software Engineering Daily.

Building a monitoring system is a complex distributed systems problem. Events are produced from different points in an application and must be aggregated in order to form metrics. These events are often ingested by a time series database, which forms the backbone of our monitoring system. Theo Schlossnagle is the CEO of Circonus, where he has been working on architecting the company’s monitoring software for six years. In this episode, we talk about how to build a monitoring system and the requirements for the underlying time-series database, as well as what monitoring even is. The post Monitoring Architecture with Theo Schlossnagle appeared first on Software Engineering Daily.

In order to move software updates from the development team to production, companies do a variety of things. Some teams might email files to each other or use FTP or even floppy disks. Most companies today at least use version control systems like Git together with separate servers for development and production. When code is ready to move to production, a build that is on the development server gets copied over to the production servers, and the production servers begin serving real users. This process is known as deployment, and over the last few decades companies have started deploying more rapidly (even “continuously”), leading to faster iterations and better feedback loops between the software development team and the users of the product. A particularly effective version of this workflow is known as continuous delivery. In today’s episode, David Rice from ThoughtWorks joins the show to give a short history of continuous delivery, and how continuous delivery actually looks in practice.   The post Continuous Delivery with David Rice appeared first on Software Engineering Daily.

Kafka Streams is a library for building streaming applications that transform input Kafka topics into output Kafka topics. In a time when there are numerous streaming frameworks already out there, why do we need yet another? To quote today’s guest Jay Kreps “the gap we see Kafka Streams filling is less the analytics-focused domain these frameworks focus on and more building core applications and microservices that process data streams.” Jay is the CEO of Confluent, a company that is building Kafka technology, and he is one of the original authors of Kafka. We began with a brief history of why Kafka evolved to be the message broker of choice for so many data engineering stacks, then we got into a discussion of streaming–what is streaming? How does Kafka communicate with streaming frameworks? Our conversation concluded with a discussion of Kafka Streams and the future of the Kafka data platform. The post Kafka Streams with Jay Kreps appeared first on Software Engineering Daily.

Platform as a service can mean different things to different people. The most prominent feature of a PaaS is the ability to abstract away issues that every developer within an organization has to deal with. As an example, developers today don’t need to fear scalability and load balancing issues as much as engineers of the 90’s and early 2000s. We can develop our applications without worrying about the scaling that is going on under the covers. Sinclair Schuller is the CEO of Apprenda and he joins the show today to discuss what is involved in building a PaaS, particularly one targeted at enterprises. His company was started in 2007, just a few years after AWS got started. We also talked about the potential Docker Fork, for which Sinclair had useful business and engineering perspectives. The post Platform as a Service with Sinclair Schuller appeared first on Software Engineering Daily.

Google builds cloud services for developers, such as PubSub, Cloud Storage, BigQuery, and Cloud DataStore. On Software Engineering Daily, we’ve done lots of shows about how these types of services are built. In this episode, we are zooming in on the interaction between the developer using a cloud service and the design and engineering of the client APIs. To build a useful cloud service, Google has to make a common way of interacting with that service from any programming language on any device. Jon Skeet is a longtime engineer at Google, and he joins the show to explain how Google uses generated code to make the creation of those APIs more streamlined. We talk about gRPC, protocol buffers, C#, and lots of other topics. The post Cloud Clients with Jon Skeet appeared first on Software Engineering Daily.

Batch and stream processing systems have been evolving for the past decade. From MapReduce to Apache Storm to Dataflow, the best practices for large volume data processing have become more sophisticated as the industry and open source communities have iterated on them.   Dataflow and Apache Beam are projects that present a unified batch and stream processing system. A previous episode with Frances Perry discussed how they work in detail. In today’s episode, Eric Anderson discusses Cloud Dataflow, a service from Google that lets users manage their data processing pipelines without having to spin up individual servers to run them on.   Cloud Dataflow–like the “serverless” movement we have done several shows on–represents a growing shift towards cloud providers offering services that abstract away the operational challenges of managing compute nodes. If you have suggestions for topics in this area, please do send me an email. The post Cloud Dataflow with Eric Anderson appeared first on Software Engineering Daily.

Slack is a chat application that is rapidly growing in popularity. The focus of Slack is to create a polished, responsive tool for productivity that cuts down on the emailing, context switching, and useless meetings that take place at a typical enterprise.   Keith Adams, the chief architect at Slack, joins the show to explain how those high level principles translate into engineering decisions.   Keith previously worked for 7 years at Facebook, contributing heavily to the tools that make PHP easier to develop with at scale. Slack’s core product is also built with PHP, and Keith discusses the similarities and differences between scaling Facebook and Slack. The post Slack’s Architecture with Keith Adams appeared first on Software Engineering Daily.

Uber has a software architecture with unique requirements. Uber does not have the firehose of user engagement data that Twitter or Facebook has, but each transaction on Uber is both high value and time-sensitive. Users are paying for transportation that they expect to be available and reasonably close by. When Uber’s system is trying to match a rider with a driver, availability is favored over consistency. It is important that the rider can always get some driver, even if it is not the best driver. Ringpop is a system built at Uber to provide scalable, fault-tolerant, application layer sharding. Ringpop consists of a membership protocol, consistent hashing, and forwarding capabilities. Jeff Wolski is a software engineer at Uber working in Ringpop, and he joins the show to explain how Ringpop brings coordination to distributed applications. The post Uber’s Ringpop with Jeff Wolski appeared first on Software Engineering Daily.