System architects must be good story tellers, because we picked up a lot of interesting tales of system design — both good design and design gone awry — at the O’Reilly Software Architecture conference. View in browser »
The New Stack Update

ISSUE 106: Microservices Comes of Age

Talk Talk Talk

“Caches can make good things better, but they can’t make something wrong good.”

Allard Buijze, AxonIQ CTO, at the O’Reilly Software Architecture conference, New York.
Add It Up
47% of Surveyed Companies' Employees Are Open Source Contributors
Get Ready, Get Set, Go: Survey Recap
A 2017 survey of over 6,000 Go language (Golang) developers found satisfied developers increasingly using it for their work. Yet, as they fine tune their use cases, developers are more likely to say a lack of critical features is inhibiting Go’s adoption.

Overall, 64 percent strongly recommend Go and 58 percent strongly prefer using Go for their next project. Although Go fans may be over-represented in the sample, the finding is confirmed by Stack Overflow’s 2017 Developer Survey. Compared to most other languages in that report, Go users were more likely to express an interest in continuing its use.
The percentage of developers using Go for work rose to 67 percent of respondents as compared to 62 percent in the 2016 version of the survey. At the same time, those writing API/RPC services in Go rose five percentage points to 65 percent compared to last year.
With more people using Go for work projects, citation of several challenges increased, with lack of generics doubling, bringing it to #2 on the list. The top challenge mentioned is lack of dependency management. Even more illuminating are the reasons Go is not being used on more projects. Those saying Go is inappropriate for the project dropped from 24 percent in 2016 to 16 percent in the current study. At the same time, those saying a lack of critical features is impeding additional use jumped from 11 to 19 percent.
What's Happening

GitOps is an iteration of DevOps as applied to cloud-native and, in particular, Kubernetes, which has a strong emphasis on declarative infrastructure, said Weaveworks CEO Alexis Richardson in this episode of The New Stack Makers. Richardson said the Weaveworks team got excited about the concept of GitOps through the ongoing management of Weave Cloud, the company's deployment, monitoring and management SaaS for containers and microservices.

GitOps for Kubernetes: A DevOps Iteration Focused on Declarative Infrastructure

Microservices Comes of Age

As a demographic, system architects must be good story tellers, because we picked up a lot of interesting tales of system design — both good design and design that have gone horribly awry — at the O’Reilly Software Architecture conference, held in this week in New York.

The hot topic was, of course, microservices, even if most of the users we spoke with were a long way from running a microservices architecture. They were peeling bits of microservices from their monoliths and came to the event with questions about the issues they were contending with.

In his keynote, Chris Richardson, creator of the original Cloud Foundry software and noted author on the subject of software architecture, encouraged the crowd to think of a set of microservices as an event-driven system. Any change in an application is, in fact, an event, and if that application communicates that event, another application could ingest that information for use in its own event-generation.

This explains why the Kafka stream processing platform is quickly becoming a cornerstone of many highly-scaleable web operations. We learned from Netflix’s Steven Zhen Wu how that company’s data processing pipeline ingests about two trillion events/day (~2PB). The company created a stream-processing-as-a-service that offers devs templates and ops support for easily creating custom jobs.

Of course, an event-driven architecture brings up a range of problems. One is reliability. If your server is waiting for 20 items from other components before returning a web page, the user may complain of pokey performance. So you have to move towards a model where the services provide info to the web server as a push service, rather than a pull one, noted Cornelia Davis, Pivotal senior director of technology, in her talk.

Another issue is complexity. How do you keep of track of these services, and debug them when something goes wrong? About this issue, we spoke with Bernd Rucker, one of the creators of a state engine called Camunda, which provides a way to coordinate microservices to show developers exactly what went wrong where.

There were a couple of other themes at the conference. One is growing resentment of configuration files. Netflix’s Nora Jones talked about how she sees more operational issues come from configuration errors than from actual code. Warby Parker’s Robert Leftkowitz, in a great talk about technical debt, discussed how it is possible to do away with configuration code altogether, by moving the configuration details into the code itself, or insert during runtime.

Keep an eye out next week or so for stories on all these topics. Only on The New Stack, of course!

Mesosphere’s Data Center Operating System Will Soon Offer a Managed Kubernetes

For the pending release of version 1.10 of the Data Center Operating System (DC/OS), Mesosphere is adding full support for Kubernetes, alongside its own container orchestration system Marathon. This is also the first release to support the Universal Container Runtime, allowing developers to use Mesosphere’s tools to create Docker images, store them in a registry, and run them on DCOS. For Moving Fast and Not Breaking Things takes on the challenges of continuous delivery. The technology consists of two parts: smart automation and continuous verification. Instead of scripting, Harness uses a model to define what you want to achieve. The user interface allows developers to model a pipeline by drawing lines between tasks that the system will automate.

Off-The-Shelf Hacker: Before ‘Open Source’ There Was Unix

We all celebrated the 20-year anniversary of open source a couple of weeks ago. While open source has certainly been a positive part of Dr. Torq’s tech career, he delves even further back into his own personal history, finding that hacking and *nix (Unix, Linux, etc.) has had a slightly bigger effect on him. Take a trip down memory lane with FORTRAN, Sun Microsystems, early Linux distributions like Slackware, the green-glowy VT100 monitors and even hot rods.

On The Road
Cloud Foundry Summit // THURSDAY APRIL 19, 2018 // BOSTON CONVENTION CENTER, ROOM 254-B


Cloud Foundry Summit
Join us in Boston for the first stop on the 2018 TNS Pancake Breakfast world tour as we discuss how Cloud Foundry and Kubernetes are growing up. How do developers and users spend their time now that open source software projects are giving them such a hand in building applications at scale on open common platforms? That’s just a taste of what we’ll chat about over carbs and conversation at the Boston Cloud Foundry Summit. 20% off with code CFNA18TNS. Register Now!
FREE EBOOK: Learn about patterns and deployment use cases for Kubernetes.
The key to successful deployment of Kubernetes lies in picking the right environment based on the available infrastructure, existing investments, the application needs and available talent. Depending on whether Kubernetes is deployed on premises, on a single cloud provider, hybrid cloud or multi-cloud, users will face different technical challenges and will need a different set of tools for deployment. These factors also affect how operations teams approach security with Kubernetes, and it’s critical to understand security in the context of these environments.
Download The Ebook
We are grateful for the support of our ebook foundation sponsor:

And our sponsors for this ebook:

Copyright © 2018 The New Stack, All rights reserved.

Want to change how you receive these emails?
You can unsubscribe from this newsletter