devopsy.kinsta.cloud

Tag: @TechFieldDay

Operations, Code or Die

There are numerous definitions of coding ranging from developing entire applications in Java or C++ to scripting in PowerShell to editing YAML configuration files. But people rarely believe that any of these is an essential skill for the operations teams.

Panelists at the recent Cloud Field Day event anticipated during the Delegate Roundtable conversation that soon artificial intelligence (AI) would take over coding, making the primary skill, the ability to articulate requirements to AI. They acquired their understanding by working with physical hardware before virtualization, followed by writing infrastructure automation. However, they do not think newer or future IT professionals will take the same route.

Eric Wright, chief content officer at GMT Delta, who spoke of his experience writing his first infrastructure code, said, “Writing that code taught me to think differently about the problem.”

The fundamental change in approach usually moves to a declarative approach: Writing a file describing how a system should be configured and using a software tool such as Ansible to ‘make it so.’ The description is usually a configuration file written in a specific format for the tool; Ansible uses YAML-formatted text files. When we must change the system with a declarative approach, we change the description and use the software tool to ‘make it so’ again. The description becomes the source of truth, and these descriptions are stored in a source control repository, the same way that the application’s source files are maintained. The prior approach was imperative, where a series of instructions are written to achieve a state rather than describing the state.

Eric and I both had experience with physical servers before virtualization and moved on to cloud infrastructure as that trend emerged. Our experience of these changes shaped how we view current technologies and innovations. There were no declarative tools for deploying a rack full of physical servers or upgrading to a new version of Windows. Declarative tools were new, and we learnt to use them.

More recently, many people entering the IT workforce have probably started with declarative tools and authoring configurations under version management for deploying onto public cloud platforms. These professionals do not need to understand server hardware assembly or rack cabling as they consume resources through software. The rites of passage for my generation of infrastructure and operations professionals are as irrelevant to them as my father’s mini-computer experience of the 1980s is to me.

But coming back to whether coding is required or if AI tools are good enough to create these declarative source files, the consensus was that the current generative AI tools such as ChatGPT-4 require careful prompting to produce usable configurations. Jon Myer, content creator and podcaster, asked ChatGPT to create an AWS CloudFormation template for a simple three-tier application. The resulting template was well-formatted but incomplete. He had to add more instructions for it to provide a usable template. The time and effort required for prompt engineering were nearly the same as those required for manually writing the template.

Other panel members had similar experiences of needing to know how to describe requirements in detail before ChatGPT could create a usable template.

Technical educator Ned Belevance described a near future when generative AI tools would improve and more abstract descriptions would suffice: “Deploy a website for me” is a more abstract prompt and removes the need to code. The real abstraction is when the prompt is a business requirement: “Build me a system to increase sales of my widgets in Japan by 30% within six months.”

As Eric advised, for our generation of professionals learning source management and writing infrastructure as code (IaC) is valid, but the current and next generations of IT professionals begin their careers with coding. They are adopting AI as a core part of their toolkit. None of us expect to be replaced by AI, but we are all at risk of being replaced by someone who knows how to use AI to be more efficient. Maybe, after all, we should be learning to AI or die.

March 11, 2025
SOUTHWORKS – Cloud-Agnostic Platforms Will Drive New Possibilities in Multi-Cloud

Cloud customers are flocking towards a new approach as the classic option of relying on a single go-to provider is falling out of favor. Multi-cloud, the concept of using two or more cloud platforms to run workloads, has emerged as a hot alternative – and a smart business strategy.

“Ten years ago, when the cloud was starting, everybody was like pick your vendor and everything will be fine, and we thought it would be,” remembered Johnny Halife, CTO of SOUTHWORKS, a company that provides the engineering horsepower required to make a go of the multi-cloud play.

“So companies invested a lot in learning the ways of AWS, Azure and GCP. What they are finding out is that just one won’t do,” he said.

Halife urges customers to have an open mind rather than “picking sides” when it comes to the cloud.

“We need to stop looking at the world as black and white, as Azure or AWS…and understand that at this point, there might be a better offering on Azure and if that’s the case, let’s build the bridges, let’s build the network, and let’s build a blueprint for that to happen.”

Being on multiple cloud has undeniable advantages. For example, it gives one room to selectively put applications on platforms that are best suited for them, giving them the perfect environment to thrive.

Or as IT is scaling beyond data centers, and edge computing is rising worldwide, being on multiple infrastructures gives one the flexibility to grow and embrace new innovations.

Cost-savings are an obvious reason as well. In the past few years as cloud costs have surged manifold, enterprises are seen trying new flavors of cloud in pursuit of lower expenses.

And with the right toolbox and a solid roadmap, one could easily be set up for success in the multi-cloud. Except there is a catch.

“This modernized multi-cloud strategy is creating a “virtuous circle” because what we perceive as benefits like cost reduction, better performance, better or different revenue streams, or even more modern and robust infrastructure are what they ask.”

But what they get may not be the same thing. Switching platforms fast and loose comes with heaps of risks.

“Full-fledged migrations or hybrids scenarios are really valuable, but they have these problems,” he said.

Due to architectural differences between platforms, an application cut out to take advantage of the discrete services of one cloud infrastructure may not be able to profit from the services of another to the same extent. Depending on its design and build, the platform will influence the application’s behavior, either positively or adversely.

In its experience of working with many startups and ISVs, SOUTHWORKS has witnessed the potential pitfalls and disasters.

“When I talk with companies that have already gone through from on-prem to cloud, or from this stack to that stack, usually they say it was painful, and a lot of them messed up on training the people.”

Building cloud-agnostic applications can help get ahead of this problem and tap the true value of multi-cloud. Such applications can safely ignore the environment they’re running on, eliminating the need to align to one particular platform.

SOUTHWORKS puts its deep experience and familiarity with all of the major cloud providers to help design multi-cloud solutions for clients that allow their applications that formerly ran on separate platforms, to be in multiple places at once, while supporting the core business requirements.

The neutrality introduces broad interoperability and cloud resiliency, while also giving enterprises a chance to fairly weigh their options when it comes to workload balancing.

SOUTHWORKS’ cloud adoption, migration and app modernization solutions are designed to help customers get through the transition frictionlessly, and allow leveraging the best offerings of each provider. The scope of their work, based on customer stories shared at the AppDev Field Day event in November at KubeCon, includes shifting workloads wholesale from one cloud platform to another, turning traditional infrastructure into SaaS, and streamlining scattered technologies into a unified infrastructure.

As a systems integration company, SOUTHWORKS serves as a cloud-agnostic technology partner guiding companies towards the most suitable cloud platforms based on business case. When a customer comes to them with an open mind, they are shown all the options that work for them, Halife told.

“It’s part of our value. We don’t want to be cornered as the AWS shop or the Google Shop or the Microsoft shop when it comes to cloud,” commented Halife.

Watch SOUTHWORKS’ presentations from the AppDev Field Day event at KubeCon for a closer look at their services.

December 5, 2024
Heroku Uplifts Developer Experience with a Shorter Build and Deploy Cycle

Across the country, companies are on a quest to find an easy button to fast-track delivery of applications to customers. An accumulating number of side tasks are continually pushing out their time to market.

A central issue that has plagued app development companies is the stretched-out phase of deployment that begins after developers have written the code, and the lifecycle management process thereafter.

“It’s really 3,000,000 steps that you need to go through to get the code out to your customers,” said Adam Zimman, sr. director of product marketing at Heroku.

From building to scaling, the myriad steps and chain of decisions that make up the journey, amount to a lot of secondary tasks that consume time and resource from the developers’ routine. Most of it is manual, and developers liken it to flying blind with no safety harnesses.

Confusingly, industry averages will have you believe that the length of deployment is measured in hours, but most teams do not stack up to that. For the majority of the industry, deployment drags on for weeks. With too many moving parts, frequent needs for change, and no clear milestone, developers have to walk a tightrope.

A fitting solution, they say, is to use automation to keep the process closed, quick and painless.

A Salesforce company since 2010, Heroku has designed a developer-friendly Platform as a Service (PaaS) solution for creating, deploying and managing applications that it says uplifts developer experience from poor to “extraordinarily pleasurable”.

The platform makes a quick work of launching applications into the market bringing up companies to be first-to-market players.

Built with the goal to end the need for developers to get into the weeds of infrastructure plumbing and app maintenance, Heroku takes over the grunt work that slows down innovation and delays release of new applications. This leaves them in full control of ideation and coding, the real work.

“Heroku is that platform that engineers go to when they need things to work, and don’t want to think about it,” said Zimman while presenting at the AppDev Field Day event in California, “because we take care of all of the scaling, making sure that the apps stay up and running.”

Under the hood, the solution offers a breadth of functions that include integration, lifecycle management, automation, configuration, provisioning and autoscaling – to enable this. The sweeping abstraction gets customers “from app idea to customer experience as quickly as possible with as little effort as possible,” he said.

Heroku is a “very opinionated system. We think about all of the abstractions that you need to normally consider when bringing an application to market.”

An example is cloud migration. Companies struggle migrating workloads to public cloud platforms because of the change it entails. “Heroku is a great way to shortcut that lifecycle and be able to migrate quickly and easily,” said Zimman. “You don’t have to worry about understanding the 7000 plus services that you can get access to on AWS. Instead with Heroku, you take advantage of the primitives that you need, and we take care of the rest.”

Heroku is a “polyglot platform”, meaning it can support applications written in different programming languages the same way. There is no need to make changes to run them on Heroku, say its makers. If it’s written using a supported language or framework, Heroku will work it out.

Here’s everything that Heroku does under the hood to make lives easy for developers, and shorten the time-to-market for companies.

Applications are deployed on Heroku via a Heroku app which is essentially a second instance of the app codebase hosted on Heroku. Users are required to introduce the source code from a standard Git repo. As Heroku receives the app source, it initiates a build. Code is compiled, dependencies are retrieved, necessary assets are created, and finally, everything is assembled into what is known as a slug.

This slug, featuring all of the fundamental pieces, is a pre-packaged copy of the application that is ready to run. Heroku executes the slug on a dyno which is an isolated Unix container environment designed to execute code based on user-specific command.

A user can choose to run any number of dynos at a given time from the dyno manager. As a new version of an app is deployed, existing dynos are killed off and substituted with new ones, keeping the current formation intact.

All dyno activities, and logs from the routers and other platform components, are streamed into Logplex, a system that records entries in real-time. Applications can be easily scaled by increasing the number of dynos running.

Configurations sit outside the code and can be tweaked independently from the source code. They get stored in configuration vars from where they are programmatically pulled down at runtime.

Zimman brought a round of projected numbers to underline the developer efficiency gains. Not having to worry about or learn all of the ways different components fit together, may lead to a jump of up to 40% develop productivity, and 30% reduction in developer expenses, he said.

“We’re providing a prebuilt template that they’re able to use to deploy applications much more quickly.”

A growing marketplace of add-on services featuring over 500 add-on and build packs offers customers ready-to-use capabilities. Messaging, backup, search, database, uptime alerts, email services – users can choose and attach them to the applications as required.

Heroku is commonly mistaken for a product designed for startups, but in its current form, the product no longer fits that mold, said Chris Peterson, sr. director of product management.

“In the last few years, the company has worked to make sure that Heroku scales as the business scales.”

While the platform adapts to the changing business requirements, users are in control of the resources they want to use and can add features as they go.

Developed in 2007, more than 13M apps and 38M managed datastores have been built on the Heroku platform till date.

A significant number of Salesforce apps run on the platform. “We have a lot of very deep integrations to extend the Salesforce ecosystem quickly, easily and seamlessly, so that people can build their Salesforce apps with Heroku.”

Learn more about Heroku by watching their presentations from the AppDev Field Day event.

November 26, 2024