Decades of building infrastructure on the customer side, and the forces I kept running into: costs that stopped making sense, regulation that stopped being optional, corporate politics that outweighed the architecture, lock-in that cost more than anyone modelled, and now the hard physical limits of AI-driven demand. What those forces mean for CloudStack, seen from the outside and why I decided to step inside.

The era of “Cloud First” is ending. Driven by accelerating regulatory fragmentation, unsustainable economics, and geopolitical tension, organizations worldwide are pivoting to a new mandate: sovereign, pragmatic infrastructure. But as data localization laws multiply and cross-border restrictions tighten, how do enterprises maintain innovation without surrendering control?

In this talk, we move past the hype to examine why open infrastructure is no longer just a technical alternative – it’s a strategic imperative. Drawing on decades of hands-on cloud architecture experience and unique governance insights from serving as both an Apache CloudStack PMC member and former Vice President, this session explores how neutral, open-source IaaS is becoming the global standard for compliance, resilience, and interoperability.

For two decades, cloud platforms have been optimized for enterprise IT. The next decade will be defined by something entirely different: AI-native applications, accelerated computing, distributed architectures, and open ecosystems.

As organizations rethink where and how workloads run, they are looking beyond hyperscalers toward platforms that offer flexibility, transparency, and innovation without lock-in. This keynote explores the technology and market forces driving that shift, why open-source communities are becoming strategic innovation engines, and what the CloudStack community can do today to shape the next generation of cloud infrastructure.

This session provides a vision for the future and challenges attendees to think differently about the role they can play in building it.

The next major CloudStack LTS version will be 4.24 and is expected by the end of 2026. The CloudStack 4.24 Release Manager (RM) is Rositsa Kyuchukova and co-RM Nicolas Vazquez. Rositsa is a committer and Nicolas a committer and PMC member of the project.

This session will give a walkthrough the new features and enhancements expected for the CloudStack 4.24 LTS version.

Facing rising licensing costs and growing concerns over vendor lock-in, a major retailer made the bold decision to move its entire virtualization estate, thousands of production VMs, from VMware to Apache CloudStack. In this customer testimonial, they share their journey first-hand: why they chose CloudStack, how they built the business case, and what it really takes to migrate at scale without disrupting critical retail operations.

Generic LLMs are trained as much in the works of Shakespeare (if not more) as they are in Apache CloudStack. They can talk about clouds, but they will still hallucinate API parameters, misread agent failures, and declare success while the infrastructure is on fire. Understanding CloudStack is not only a model problem, it is a **harness** problem: playbooks, tools, docs, and escalation paths that force correct behavior.

This talk shows two ways to improve that scenario. We build harnesses that encode real ACS knowledge (Markdown skills, labs, APIs, log sources, vector embedding, AI-docs, etc…) so agents explore, validate, and act instead of guessing. On top of that, we explore how to train specialist agents with tool discipline and domain knowledge, who are able to properly understand ACS, how to document issues and escalate those to more generalist SOTA agents in a well harnessed form.

Join the talk to discobver an end-to-end walkthrough of the NAS Backup and Recovery workflow in Apache CloudStack, covering backup creation, management, and restoration.

It also explores recently introduced capabilities, including extended primary storages support, cross-zone restore and incremental backups.

CloudStack Extension for efficient, immutable and fast global deduplicated Backup with individual GFS-Domain settings. And Restore Wizard and Reportings for Domains.

Organisations moving production workloads from VMware to Apache CloudStack commonly treat migration, disaster recovery and ongoing backup as separate projects, often with separate tools — creating operational complexity, and leaving historic recovery points tied to the platform they are trying to leave.
This session explores a different model: one protection platform across the complete workload lifecycle — protecting workloads on VMware, replicating them into Apache CloudStack, supporting controlled failover, and continuing to protect the same workloads in production on CloudStack, preserving each workload’s backup lineage across both environments rather than starting a new protection history after migration.

The approach is grounded in production experience. In one recent deployment, a UK organisation with fewer than 30 days to make a VMware licensing decision deployed the Sendense replication appliance into its VMware environment, began replication into a CloudStack-based cloud in under an hour, and migrated its complete VMware estate in 14 days — with protection continuing after cutover and ongoing Backup as a Service.

For cloud providers, this workflow is also a commercial model. Replication that starts within the first hour shortens mean time to revenue on a migration engagement; the platform then turns a one-off migration into recurring BaaS and DRaaS revenue, with deduplicated EBA storage improving margin on every protected terabyte; and the customer keeps continuous protection and a complete recovery history throughout the move — a better customer outcome.

This session introduces the NetApp ONTAP plugin for Apache CloudStack and explores the storage capabilities it brings to KVM environments. We will discuss plans to support primary storage pool-specific provisioning workflows, enabling greater control over how storage resources are allocated and managed. The session will also cover planned support for local VM snapshots and CloudStack volume snapshots, providing an overview of the integration, its key features, and its future development.

CloudStack has encrypted volumes on KVM since 4.18, but the key for every encrypted volume sits in the management server database and cannot be rotated.

4.23 adds a Key Management Service: the volume key is encrypted by a master key that never leaves an HSM, and CloudStack stores only the encrypted form.

Rotating the master key does not re-encrypt any volume data, and volumes encrypted before the upgrade can be migrated.

I will cover how volume encryption works today, what changes with KMS and what it does not cover, then demo it using a simulator.

In this session, Matthias and Paul will show you how to easily deploy open-weight models on your private infrastructure and how to pilot everything from Apache CloudStack

New to KVM or evaluating CloudStack as a cloud orchestration platform?

This interactive workshop gives every participant access to a live CloudStack environment, with no installation or preparation required. Working in your own tenant, you will explore CloudStack’s core capabilities by performing the same day-to-day tasks administrators use in production.

You will create virtual machines from templates, configure networking, resize resources, perform live migrations, create snapshots and clones, and see how CloudStack handles common operational workflows.

Along the way, we will explore the practical questions teams typically have when getting started with CloudStack:

• How are compute, networking, and storage managed?
• What does day-to-day administration look like?
• How are common virtual machine lifecycle operations performed?
• Which capabilities should be evaluated during a proof of concept?
• What should teams new to KVM expect when operating the platform?

The workshop is designed as a practical introduction for anyone evaluating CloudStack or looking to better understand how it works in a real environment. You will leave with hands-on experience, greater confidence in the platform, and a solid foundation for building your own proof of concept.

A marquee feature is coming to Apache CloudStack: a Disaster Recovery (DR) framework that brings capabilities inspired by leading platforms such as VMware Site Recovery Manager. Designed for KVM hypervisors with Advanced Networking, the framework is being developed through the joint efforts of ShapeBlue and GTT.

The DR Framework will integrate with third-party storage replication technologies, allowing storage vendors to connect their own replication solutions through dedicated plugins.

Natively integrated into Apache CloudStack, it will enable administrators to run disaster recovery simulations, validate recovery procedures and manually trigger failover operations directly through CloudStack.

Apache CloudStack continues to strengthen its integration with Ceph, delivering new storage capabilities and operational enhancements for cloud operators.

This talk explores recent developments including improved RBD storage support, snapshot and template management optimizations, and streamlined lifecycle operations for Ceph-backed volumes.

We will discuss architectural improvements that enhance reliability and performance, as well as new features that simplify day-to-day storage operations at scale. Attendees will gain practical insights into deploying and managing Ceph as storage backend in CloudStack environments.

Migrating a VM from VMware to KVM sounds simple — read the disks, convert them, write them out — but building a pipeline that is efficient, scalable and low-downtime takes considerably more engineering. This session walks through the architecture of a warm VMware-to-KVM migration: an initial full copy, iterative synchronisation of changed blocks, and a short final cutover. We will compare the roles of VMware VDDK, nbdkit, qemu-img, qemu-nbd, libnbd and nbdcopy, and how the chosen data path affects sparse data, changed-block processing, transfer performance and operational complexity — alongside VDDK transport modes, snapshot consistency, worker placement, parallelism, failure recovery and orchestration. A further focus is destination-storage abstraction, since the optimal write strategy differs between file-based qcow2, Ceph RBD, LINSTOR and other raw block targets, and a migration engine should support them without coupling its core data path to one platform. Apache CloudStack serves as the practical orchestration example — though the concepts apply to any KVM/libvirt environment — and the talk, based on hands-on implementation and performance testing, closes with a demonstration of an iterative warm-migration workflow.

This workshop demonstrates a complete workload protection lifecycle live, using real infrastructure rather than slides alone. A Sendense Node Appliance runs in a remote VMware environment; a Sendense Hub Appliance runs in an Apache CloudStack environment; and backup data is written to Enterprise Block Archive (EBA). The virtual machines, backup jobs and replication jobs shown are real.

Attendees will see the remote VMware environment onboarded into Sendense; VMware VMs discovered and selected for protection; backup jobs written into EBA, with EBA Smart Transport operating on the backup data path; and recovery points created from the VMware source environment. The same VMs are then configured for replication into Apache CloudStack, failed over, and shown running on CloudStack. New backups are taken from the workloads after the transition, and each workload’s backup lineage shows its recovery history from both VMware and CloudStack as a single continuous record. The session closes with automated boot evidence (recovery validation) and granular file recovery from a protected VM.

The aim is to show that backup, replication, disaster recovery and migration are different operations within one protection platform, rather than isolated products. This is a practical technical session intended for CloudStack operators, MSPs, cloud providers, architects and infrastructure engineers. Pre-created recovery checkpoints will be available as sensible fallbacks should venue connectivity or replication timing prevent an individual operation completing during the session — the demonstration itself is live.

Learning outcomes

• Connecting VMware and CloudStack environments to a single protection platform.

• Sendense backup architecture and Enterprise Block Archive (EBA), including EBA Smart Transport.

• Replication, recovery and failover into Apache CloudStack.

• Continuing protection after a platform transition, and how backup lineage is preserved across source and destination.

• Recovery validation through automated boot evidence, and granular file-level recovery.

• What the workflow means for providers delivering BaaS and DRaaS on CloudStack.