J.B. Hunt Reduces Developer VDI Costs by 90% with Coder

Overview

J.B. Hunt Transport Inc. is one of the largest transportation and logistics companies in North America, ranking among the Fortune 500. Their mode-neutral approach generates value for their customers by eliminating waste, driving out costs, and enhancing supply chain visibility. J.B Hunt combines one of the largest company-owned fleets in the country with extensive third-party capacity available through its J.B. Hunt 360™ digital freight marketplace. To drive efficiency and deliver value for their customers, J.B. Hunt invests heavily in their people, technology, and capacity. Technology that empowers is a core foundation of the company. That’s where their developers play their part. They provide the information systems to keep things rolling efficiently and dependably. Daniel Trimble is the Director of Site Reliability Engineering (SRE) at J.B. Hunt. He and his team manage the infrastructure for J.B. Hunt’s software systems, including the environments that their developers work in.

Challenge: Virtual convenience has its costs

J.B. Hunt’s developers work on solutions to manage the logistics of shipping and transportation. The three dozen product teams use the same stack, which includes Docker, Java, and .NET Core. Most of their projects are implemented as micro-services. Each team has its own dependencies to satisfy the product’s needs. Of the 700 developers at J.B. Hunt, about 150 are remote. The simplest way for the SRE team to provide each remote developer with a development environment was through virtual desktops in the cloud accessed by Remote Desktop. J.B. Hunt used Azure’s Virtual Desktop service to host these environments. For version control and devops, they use Azure DevOps.

This convenience came with trade-offs though:

• Remote developers reported laggy user interfaces, made worse by going through a VPN.
• With a VM for each developer, inconsistencies crept into their setups. With each team documenting their own development environments equally inconsistently, onboarding took a while, up to 2 weeks.
• Not surprisingly these variations in dev environments increased support costs for the SRE team. It also made developer workflow susceptible to “works on my machine”.

But the highest direct cost for this convenience was cloud spend. Typically, remote developers worked with a 4-CPU, 16 GB VM, and sufficient storage. That’s among the highest-priced cloud resources there is. Now multiply that by 150 developers, plus licensing for Microsoft Windows, 8 hours/day, 5 days/week.

Solution: Lower costs, more flexibility

The SRE team set out to find a way to reduce costs. They first explored a mid-market CDE but gave up when they discovered its lack of support for Azure DevOps.

The team then discovered that Coder doesn’t impose this constraint because it takes a provider-agnostic approach. If you can provision your development environment with a template, you can manage it with Coder. This flexibility appealed to the SRE team.

First, they specified a base development environment and containerized it. To manage these environments, they set up Coder to run in a GKE cluster, where each developer gets a pod. Developers access their remote development environment from a locally-running instance of VS Code.

With this setup Coder and Kubernetes handle resource management as developer demand dictates. The SRE team no longer has to worry about idle VDI instances. With a standardized dev environment, there is no need to reinvent the wheel for devops. The SRE team is able to maintain their investment in Azure’s DevOps services.

To onboard developers, the SRE team’s adoption plan includes an internal bootcamp for remote and new developers. Developers learn how to log in to Coder to start workspaces and how to use a new IDE, VS Code. They also learn to re-map their Windows command-line skills to Linux.

Results: Lower cost, better developer experience

With about 150 remote developers now using Coder instead of virtual desktops, the SRE team is able to regain control of remote development. The most impressive benefit is the reduction in cloud spend. The combination of smarter resource management and less expensive cloud services made J.B. Hunt executives happy to see developer VDI cloud costs (a subset of their total cloud spend) drop to less than 10% of what it used to be.

They’ve also eliminated the inconsistencies that used to keep the SRE team busy with support issues. Coder templates let the SRE team provision reproducible, easy to support, consistent development environments. Now the team’s most common support issue is to remind developers to use a forward slash instead of the backward slash that Windows uses for path names.

Developer experience improved on a few fronts:

• Onboarding is much faster with Coder. The base template handles the complications of setting up the right versions of Docker Desktop, Java, .NET Core, and so on.
• Each developer can still customize their own workspaces for things like including unique dependencies, but there’s always a base workspace that only the SRE specifies, letting them document and support it more effectively.
• Developers don’t have to wait for Remote Desktop latency through VPN. With a locally-running, responsive IDE, they don’t suffer from user interface lag anymore. Starting a workspace is faster too.

Conclusion: True convenience

When the SRE team at J.B. Hunt hit a limit on budget and inflexibility, they chose Coder to support remote developers. Coder reduces J.B. Hunt’s cloud costs, gives developers a better experience, and lets the SRE team simplify their support efforts.

Challenges

• High cloud spend
• Heterogeneous dev environments
• Laggy user interface from remote desktop

Outcomes

• Over 90% drop in developer VDI cloud services, saving hundreds of thousands of dollars/year
• Lower support and development costs: from over 150 VDI development environments to a handful of environments on Kubernetes
• Improved developer experience

Industry

Number of developers

Related Content

• Four Reference Architectures for CDEs: Kubernetes vs. VMs
• CDEs + Enterprise Browsers: The Modern Alternative to VDI for Developers
• Virtual Desktop Infrastructure Development Environment

Related Content

• Four Reference Architectures for CDEs: Kubernetes vs. VMs
• CDEs + Enterprise Browsers: The Modern Alternative to VDI for Developers
• Virtual Desktop Infrastructure Development Environment