Hi Kata Containers community, I’m reaching out to share aggregated results to the latest Kata Containers User survey. There were 5 responses over the past 1 year, please find the anonymous and aggregated results below. Also, please note that the sum of percentages can go higher than 100 for questions where respondents are allowed to pick more than one answer. 1. General information and feedback Net Promoter Score (NPS) score: 8.2 NPS score feedback: * It works and solves a very important use case for SaaS * Documentation could be improved * People want more use cases and case studies Current deployment stage: * Evaluating: 40% * In production: 40% * Planning to put into production: 20% What people like most about Kata: * Community * Documentation * Enhanced security * Support for different kernels * Customizable Anything missing or confusing in the documentation: * Yes: 60% - What people found missing: ~ How to use Huge Pages with Kata, and similar guides are missing ~ Troubleshooting guide, ie how to debug a pod VM - Great documentation, but it seems to be getting out of date * No: 40% 2. Environment Where Kata is deployed: * Public Cloud: 80% * On-premises: 20% Cluster deployment, canary deployment or not: * No, infra team will manage it: 60% * Yes, I would like for kata-deploy to do it in this way: 20% * Only using Kata for evaluation: 20% Scenarios Kata will be used for: * CaaS: 60% * Microservices: 20% * CD: 20% * Multi-tenancy: 20% * Containers at the edge: 20% * Confidential Computing: 20% Importance of below functionality (highest score is 5): * Network performance for containers: 4.6 * Disk performance for containers: 4.6 * Container launch time: 4 * Ability to use a customer kernel &/or custom rootfs: 4.8 * Container startup speed (container "boot speed”): 4 * Enhanced security: 4.4 * Memory density (maximum number of containers that can be run on a single system): 4.2 * Networking performance & latency: 4.6 * Storage performance: 4.4 Orchestrators used: * Kubernetes: 100% Workload types: * Mix: 40% * Long-running: 40% * Short-lived: 20% Trusted vs untrusted workloads: * Trusted: 80% * Untrusted: 20% Base host OS. People listed the following: * Debian * Fedora * Ubuntu, RHEL/Alma/Rocky * CentOS Stream * CBL Mariner Hypervisors used: * QEMU: 40% * Cloud hypervisor: 40% * ACRN: 20% Storage drivers used: * VirtIO-FS: 80% * Devicemapper: 20% * 9pfs: 20% 3. Deployment configuration Installation methods used: * Build from source: 80% * Kata deploy from K8s: 20% HW architecture: * Intel (x86_64): 80% * ARM64 (aarch64): 80% Number of instances per node: * Still in eval: 20% * Less than 10: 20% * 10 - 100: 20% * More than 100: 20% QoS class used for Kata pods: * Guaranteed: 60% * Best effort: 20% * Still eval: 20% CNI device plugins used: * SR-IOV: 40% * Calico: 40% * Cilium: 40% * Multus: 20% * MacVLAN: 20% * Kubenet: 20% Memory and CPU overhead expectations: * 60% didn’t have any * Minimal, but low CPU is more critical for 20% of the respondents * 10% is a lot, 20% is a no-go for 20% of the respondents Support required to deploy Kata: * GitHub, Mailing List, Slack: 60% * Full support: 20% * No support: 20% 4. Deployment customization Custom kernel usage: * Yes: 40% - standard kernel with the latest patches * No: 60% Custom-built Kata RootFS: * Yes: 20% — small rootfs, IO performance is critical * No: 80% Custom Kata configuration file: * Yes: 40% — to tweak hypervisor parameters or to set max CPU and memory and custom virtio_fs_exrta_args * No: 60% 5. Upgrades and releases Which major release: * 3.x: 60% * 1.x: 20% - want to upgrade * 2.x: 20% - want to upgrade How upgrades are managed: * Do the builds and deploy themselves * Compatibility and maintenance efforts are high priority * Need to sort it out still, no method yet How long to support older releases: * 12-24 months: 20% * 3 months: 20% * No preference: 60% Ideal release cadence: * A new minor release every 3 months, with a LTS release maintained for 6 months: 40% * A new minor release every 3 months, with patch releases in the previous stable: 20% * Monthly releases without a stable branch: 40% How often people are willing to upgrade: * 12-24 months: 20% * Quarterly: 20% * Every 3 months: 20% * No preference: 40% Switch to RuntimeRS: * Yes: 60% * Either depends on user demands or when it becomes the recommended default How long they are planning to keep GOlang version: * Don’t care: 80% * As long as it is the recommended default: 20% Interested in learning about observability and traction: * Very interested: 60% * Somewhat interested: 20% * Neutral: 20% 6. Confidential Computing How important is Confidential Computing: * Extremely important: 80% — either driven by general interest or sensitive customer data * Somewhat important: 20% Confidential Computing and Performance isolation features: * Interested in experimenting with micro-vm image download, and have concerns about performance problems * Kata guest verification and secure runtime/agent interface; would consider confidential VMs as opposed to containers as alternative 7. Contributing to Kata Containers * Interested in contributing: 40% Best Regards, Ildikó ——— Ildikó Váncsa Director of Community Open Infrastructure Foundation