On Monday VMware announced that they will acquire SpringSource – a move that I believe was both essential and astute.
Let’s start with essential. Sure – it’s hard to argue against VMware’s merits as a hardware infrastructure-as-a-service (HIaaS) platform for both internal and external clouds. Note that for more information on cloud platforms and definitions, Drue Reeves’ paper “Cloud Computing: Transforming IT" is a great place to start. Living at the bottom of the stack (i.e., infrastructure) can only take VMware so far. Microsoft is confident that they can methodically chip away at VMware’s market share by tying their applications and management platform to their own Hyper-V-based virtual infrastructure. Richard Jones’ post “Virtualization Wars, what can VMware learn from the past?” talks about this issue in great depth. VMware’s infrastructure focus is its strength, but at the same time its Achilles heel. VMware needs to move up the stack to keep up with Microsoft long term.
I see the move as astute because SpringSource gives VMware the right platform at the right time. Chris Haddad – with our Application and Platform Strategies Service – detailed how a combined VMware and SpringSource platform will impact application development. Virtualization (i.e., server, client, application, storage, I/O, and network) and cloud are fundamentally changing the way that both server and desktop applications are delivered. Last year I wrote about how this transformational period creates opportunity for Microsoft’s competitors such as Apple. Cloud-based application delivery (both internal and external) is equally disruptive to traditional server application delivery models. What this means is that the time to redefine application delivery and to unseat the incumbents is right now.
At Catalyst, Microsoft’s Mark Russinovich talked about how virtualization is changing the way we think about OS and application delivery. Even Microsoft knows that the monolithic OS is not the platform of the future. From a platform developer’s point of view, the hypervisor represents the new hardware. OSs incapable of leveraging the dynamic nature of virtualization simply do not give applications the dynamic capabilities they may need. For example, suppose you want to add more compute resources to a Windows application. For stateless applications, you could deploy them in a network load balanced cluster and scale them out across multiple VMs. For stateful applications, scaling out isn’t as easy (some do this better than others). If you want to hot-add an additional virtual CPU, for example, the guest OS had better be Windows Server 2008 Datacenter edition. No other edition supports CPU hot-add. Clearly, this is an area that Microsoft will address, but it speaks to the problem. The same can be said for hot added memory. If the application requires a restart in order to take advantage of dynamically added memory, such features are of little value. While we’re talking about density and cloud, we also must think about efficiency. When deployed to a single physical box, application processing efficiency may not have meant so much for tier 2 or 3 apps. Put an app on a cloud-based infrastructure that uses consumption-based pricing (thinking future) and it’s another story. The more efficiently that an application and its underlying platform process data, the less expensive it is to run in the cloud.
Finally, I don’t see VMware as grabbing SpringSource to just have the “PaaS checkbox” to line up against Microsoft’s Azure platform. VMware has to think beyond that. Getting a platform for PaaS with a large install base and developer ecosystem gives VMware instant credibility, and broadens the services that can be offered by VMware’s service provider partners. Tomorrow’s dominant cloud providers won’t just have the checkboxes, but they’ll offer innovative, dynamic management, while being highly efficient (again the fewer the CPU cycles, the lower the cost). Microsoft, Citrix, and other competitors in the cloud space also realize that the opportunity exists for new cloud-centric platforms. Having a good core platform is clearly important, but tightly aligning applications and infrastructure to handle the dynamic nature of virtualization and cloud (e.g., dynamic resource add/subtract, distributed data access and potential latency, and security) is new territory. VMware’s core leadership, including Paul Maritz and Todd Nielsen have been there before, dating back to their time at Microsoft. The next steps that VMware takes will likely define how history repeats itself.
History will repeat, but which history will it be – Paul and Todd helping build the market leading software infrastructure? Or Microsoft using tight application, OS, and management integration to thwart competitors?