In September, users trying to download VMware’s Virtual Disk Development Kit (VDDK) started running into HTTP 404 errors. Broadcom had pulled the library without communication or notice. VDDK has long been a building block for backup, disaster recovery and migration tools across the ecosystem, including many of the tools organizations use to move off VMware. Cloud service providers feel the loss most. Broadcom’s licensing changes have already raised their costs, and the vast majority of providers will lose access to the VMware Cloud Service Provider (VCSP) program in March 2027. Because CSPs run repeatable migrations across many customers, they need methods that are fast, reliable and predictable at scale.
To understand what this means for CSPs, MSPs and enterprises, VMblog spoke with Sirish Raghuram, Co-Founder and Chief Growth Officer of Platform9. Raghuram explains why he sees the VDDK removal as just one migration method going away, and why newer approaches may be better. He walks through the three methods supported by Platform9’s vJailbreak tool: VDDK-based, proxy-assisted and storage-assisted migration. The storage-assisted approach, built with partners including Pure Storage, Hitachi Vantara, NetApp and Tintri, can clone a 250 GB disk in 4 seconds and complete an end-to-end migration in about 15 minutes. He also covers the opportunities this disruption creates for service providers in Kubernetes, private AI and disaster recovery.
++
VMblog: Can you start us off by defining VMware’s VDDK and explaining its importance?
Sirish Raghuram: VDDK, the Virtual Disk Development Kit, is a disk library VMware released so that third-party tools could read and access the disks of virtual machines running on VMware. For years it has been a building block for backup, disaster recovery and migration tools across the ecosystem, including many of the tools people use to move off VMware.
VMblog: What happened with VMware’s VDDK and why is it so controversial?
Raghuram: Broadcom pulled the library without any communication or notice. In September, people simply started reporting HTTP 404 errors when they tried to download it.
It’s hard to see this as anything other than Broadcom recognizing that VDDK was being used to exit VMware for alternative hypervisors, and seeing no business benefit to them in supporting that. But users, partners and the community have depended on this library for a long time, and from a customer’s perspective, the move clearly does not help.
The good news is that it removes just one migration method. There are many other methods, and several of them are better than VDDK.
VMblog: Which companies might feel the most impact here?
Raghuram: Cloud service providers are being hit by a double whammy. Broadcom’s licensing changes have made VMware a lot more expensive for them, increasing their cost of doing business. Additionally, Broadcom has terminated access to the VMware Cloud Service Provider (VCSP) program for the vast majority of providers, effective March 2027. So many CSPs are left with no option but to find an alternative, and losing VDDK at the very moment they’re under pressure to exit VMware adds to that pressure.
Service providers also don’t migrate one environment once. They run a repeatable migration process across many customers, so they need methods that are fast, reliable and predictable at scale.
VMblog: What can cloud service providers, MSPs and organizations do if they previously relied on VDDK to migrate their customers?
Raghuram: Take advantage of the newer migration methods, which are better than VDDK in many ways.
Many of the CSPs adopting Platform9 started out with VDDK-based migration because it’s the better-known method. But they ran into problems: downtime windows were longer and harder to plan for. Those who moved to proxy-assisted migration, or where possible to storage-assisted migration, have seen dramatically faster migration windows. It reduces time, reduces cost, and improves planning and operational flexibility.
So the first question to ask any migration vendor is simple: what methods do you support that don’t depend on VDDK?
VMblog: What is out there that would allow cloud service providers to migrate off VMware safely?
Raghuram: Platform9’s migration tool, vJailbreak, supports three migration methods, and only the original one uses VDDK. We designed for this possibility because we always felt VDDK could end up where it has, and because it was never the most performant method.
Our methods include:
#1 VDDK-based migration: The original method, which copies VM disks through VMware’s VDDK libraries and network block device (NBD) service.
#2 Proxy-assisted migration. We deploy a helper virtual machine in vSphere. It mounts a snapshot of the source VM’s disk alongside the target VM’s volume and does a local disk copy. While the source VM keeps running, we take refresh snapshots and copy the changed blocks, progressively narrowing the window until cutover. There’s no dependency on the VDDK libraries or the NBD server, which makes it faster and more reliable.
But it still wasn’t as fast as we wanted. In production, a database workload with 3% or 4% daily churn can take two to three days for the snapshots and changed blocks to converge. That makes it hard to plan the cutover, when you power off the source and power on the new VM on the KVM hypervisor.
#3 Storage-assisted migration. In 2025 we added a third method, built with our storage partners including Pure Storage, Hitachi Vantara, NetApp and Tintri. Instead of copying data through a proxy VM or over the network, we delegate to the storage array’s ability to make virtually instantaneous clones. We hand the array the source VM snapshot and the target disk, and the array does the clone.
The performance impact is dramatic. With Pure Storage, a 250 GB virtual machine disk clones in 4 seconds, and we’ve seen similar results with Hitachi and others. The end-to-end migration, including powering off the source VM, performing the clone, removing VMware Tools, installing the QEMU guest agent and making the other configuration changes, takes about 15 minutes. You know exactly how long the downtime will be, you’re not waiting on changed-block tracking to converge, and customers get to use the big-iron storage they’ve already paid for.
It also fits the way Platform9 approaches migration. A big part of our value is converting an existing VMware cluster in place, one host at a time, into a Platform9 cluster. Those hosts are already connected to the storage array, so it makes no sense to add the labor, configuration complexity and cost of bringing in swing gear and moving workloads to different storage.
VMblog: Is there any good news cloud service providers can take out of this?
Raghuram: Broadcom has created a lot of pain and unplanned change, but sometimes changes like this are a blessing in disguise. If I were a CSP, I’d treat this as an opportunity to get better.
First, service providers have more options than they think. Alternative platforms, including Platform9 Private Cloud Director, now offer a viable carrier-grade, enterprise-grade alternative, and the disruption is sparking innovation and investment across private cloud.
Second, new market opportunities are emerging. Many CSP customers have wanted containers and Kubernetes, but that was never VMware’s forte, and CSPs had to cobble together disparate solutions. Private cloud platforms with a built-in Kubernetes service, Platform9 being one, let CSPs monetize higher-value services such as cloud-native platforms as part of this transformation.
Third, Kubernetes is a foundational technology for AI. Over the last five to ten years, CSPs lost many modern workloads to the hyperscalers. With AI, there is real concern about data gravity and hyperscaler costs, and that gives CSPs a renewed opportunity to win back relevance with enterprise customers through private and secure AI offerings, where the data stays in the CSP’s infrastructure and the intelligence comes to it.
VMblog: With all the changes at VMware, how should cloud service providers rethink their virtualization, cloud and especially disaster recovery strategy?
Raghuram: Use the transition to build a more comprehensive cloud offering, not just a like-for-like replacement: one platform that spans virtualization, Kubernetes and private AI, so you can offer more services and win more share of wallet from your customers.
Backup and DR remain critically important through all of this, and the great news is that the ecosystem is moving with you. Major backup and DR vendors such as Veeam, which is extremely popular with CSPs, Rubrik and Cohesity are all investing in supporting alternative hypervisor platforms, they all support Kubernetes backup, and they will support the disaster recovery capabilities needed in the AI era.
Once they go through this transformation, CSPs will find that over time they get back all of the capabilities they had, and gain a more comprehensive cloud offering that makes them more competitive with the giant hyperscale cloud providers.
VMblog: Are you recommending organizations shouldn’t use VMware anymore?
Raghuram: Every VMware customer knows for themselves whether VMware is the right long-term choice for them, and I don’t consider it our place to tell people what they should do.
What we believe at Platform9 is that private clouds are incredibly important. We’re in the business of providing a capable, powerful private cloud platform that is simpler to run and spans virtualization, containers, microservices and private AI workloads, and others in the industry are investing in that too.
So my message to VMware customers is: you have options. Evaluate them, and you’ll know what’s right for you. Larger organizations have more gravity and complexity, and migration projects can be difficult, but it’s a matter of getting it done. It is not a death knell or something to resign yourself to living with. There are great options for private clouds, and some of them will get you to a much better place than you ever were with your previous legacy stack.
