The Cost of Downgrading Build Systems A Case Study of Kubernetes

Gareema Ranjan 1 , Mahmoud Alfadel 2 , Gengyi Sun 1 , Shane McIntosh 1 · 2025

Read The Cost of Downgrading Build Systems A Case Study of Kubernetes by Gareema Ranjan 1 , Mahmoud Alfadel 2 , Gengyi Sun 1 , Shane McIntosh 1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-20041.txt.

Introduction

Gareema Ranjan and three co-authors examine what happens when a large software project gives up a fast build tool for a simpler one. Their case is Kubernetes, which moved from the artifact-based tool Bazel to Go Build. They reproduce and measure full and incremental builds from change sets made during that switch.

Bazel finishes full builds faster, but it uses much more memory and adds heavier processor load at higher parallelism settings. The authors estimate that the downgrade can raise continuous integration resource costs by a large share. Repeating the study on four other projects shows the same pattern, so the trade-off matters for large projects broadly.

Catalog record

CREATORGareema Ranjan 1 , Mahmoud Alfadel 2 , Gengyi Sun 1 , Shane McIntosh 1
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE62 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-20041

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

The Cost of Downgrading Build Systems A Case Study of Kubernetes Gareema Ranjan 1 , Mahmoud Alfadel 2 , Gengyi Sun 1 , Shane McIntosh 1 Abstract Since developers invoke the build system frequently, its performance can impact productivity. Modern artifact-based build tools accelerate builds, yet prior work shows that teams may abandon them for alternatives that are easier to maintain. While prior work shows why downgrades are performed, the implications of downgrades remain largely unexplored. In this paper, we describe a case study of the Kubernetes project, focusing on its downgrade from an artifact-based build tool (Bazel) to a language-specific solution (Go Build). We reproduce and analyze the full and incremental builds of change sets during the downgrade period. On the one hand, we find that Bazel builds are faster than Go Build, completing full builds in 23.06–38.66 % less time and incremental builds in up to 75.19 % less time. On the other hand, Bazel builds impose a larger memory footprint than Go Build of 81.42–351.07 % and 118.71–218.22 % for full and incremental builds, respectively. Bazel builds also impose a greater CPU load at parallelism settings above eight for full builds and above one for incremental builds. We estimate that downgrading from Bazel can increase CI resource costs by up to 76 %. We explore whether our observations generalize by replicating our Kubernetes study on four other projects that also downgraded from Bazel to older build tools. We observe that while build time penalties decrease, Bazel consistently consumes more memory. We conclude that abandoning artifact-based build tools, despite perceived maintainability benefits, tends to incur considerable performance costs for large projects. Our observations may help stakeholders to balance trade-offs in build tool adoption. I Introduction Build tools ( i.e. , tools that automate the process of transforming source code into deliverables) are a key development resource [ 1 ] . Developers frequently invoke builds in their local environments as they modify the codebase. Automated builds are also triggered by pull requests, build hooks, or nightly scheduled workflows. Since developers invoke builds frequently, build tool performance can boost or hinder developer productivity [ 2 , 3 , 4 , 5 , 6 ] . Artifact-based build tools, such as Bazel, Buck, Pants, and CloudBuild [ 7 ] , have emerged to accelerate builds by enabling features like remote execution and caching. Unlike traditional tools ( e.g. , Make [ 8 ] ) or language-specific ones ( e.g. , Go Build), artifact-based tools maintain full control over the build process by explicitly tracking inputs, intermediate artifacts, and outputs. By constructing complete dependency graphs, these tools determine which tasks can be run in parallel and which artifacts can be fetched from a central cache, thereby minimizing repetitive build activity. Despite these benefits, it is not uncommon for projects to abandon artifact-based tools. For instance, in prior work [ 9 ] , we found that 11.62 % of projects that adopted Bazel (a popular artifact-based build tool) later downgraded to simpler build tools, such as Make or Go Build. These decisions are often driven by concerns over maintainability, contributor onboarding, and platform compatibility rather than build performance. While prior work describes motivations for downgrades, abandoning artifact-based tools may incur performance penalties. In other words: are teams that downgrade sacrificing build speed and efficiency for maintainability? In this paper, we investigate the performance and cost trade-offs of abandoning artifact-based build tools by asking: Are there performance or cost penalties associated with build tool downgrades? To address this central question, we conduct a case study of Kubernetes—one of the most popular modern open-source systems for orchestrating containerized applications. Kubernetes is an ideal case for our study because the team downgraded to Go Build after over four years of using Bazel. During the transition, the project maintained both build tools concurrently, providing us with the unique opportunity to compare both tools in a real-world setting. We replay full and incremental builds 1 1 1 Full builds run all build commands, whereas incremental builds only invoke commands that are (transitively) impacted by changed inputs. for commits recorded before the downgrade event. Our experiments consumed approximately 3,402 computational hours—equivalent to over four continuous months of execution.Through these experiments, we address the following Research Questions (RQs): RQ 1 How does the downgrade impact build duration? Build duration is a central factor in developer productivity. Bazel and similar tools are designed to reduce build time [ 10 ] , but it remains unclear whether downgraded tools retain similar performance. We observe that Bazel builds are consistently faster than Go Build, with a 23.06–38.66 % speedup for full builds and up to 75 % for incremental builds at higher parallelism levels. RQ 2 How does the downgrade impact resource usage? Performance gains may come at the cost of increased resource usage. We measure memory and CPU usage under various parallelism settings. We observe that Bazel builds require significantly more memory—up to 351.07 % more for full builds—and exert greater CPU load under higher parallelism. Go Build, in contrast, is more memory-efficient at lower parallelism levels. RQ 3 How does the downgrade impact the costs of continuous integration? Build-related costs in CI environments can be substantial, especially for large systems [ 2 , 11 ] . We estimate these costs using resource pricing models from CI providers. We observe that Bazel builds cost 22.62–39.14 % less than Go Build for full builds and up to 75.92 % less for incremental builds. While Kubernetes offers a unique opportunity to study build downgrades in a large-scale setting, its monorepo structure and scale may limit generalizability. To mitigate this and assess the generalizability of our findings, we replicate the case study on four open-source projects drawn from the dataset of our prior work [ 9 ] . Our replication demonstrates that although the performance gap between Bazel and Go Build narrows in smaller projects, Bazel consistently imposes a larger memory footprint than its replacement. Our study implies that downgrading from artifact-based tools, such as Bazel, can introduce significant hidden costs, even in projects that made the change deliberately. While maintainability and onboarding concerns often motivate such downgrades, our results suggest that these decisions may come at a steep performance and financial cost, especially at scale. For practitioners, our findings serve as a caution against assuming that simpler build tools will scale effectively without consequences. Teams considering a downgrade should carefully evaluate performance and CI cost implications using project-specific benchmarks. For researchers and tool builders, our study highlights a growing need for build systems that strike a better balance between performance and maintainability. Bridging this gap remains an open challenge, and an opportunity for the next generation of build tooling. To our knowledge, this is the first study to combine build duration, resource usage, and CI cost to quantify the penalties of downgrading from an artifact-based build system to a language-specific one. Prior work documented downgrade events [ 9 ] , but did not measure their concrete costs. Also, we do not intend to use the term ‘downgrade’ to imply a migration to an inferior tool. For homogeneous Go projects, Go Build can simplify onboarding and maintenance [ 9 ] while providing most of the

Read the complete text