In July when VMware acquired network virtualization startup Nicira for more than $1 billion, VMware showed that it’s serious about being as big of a player in the cloud’s future as it has been with the cloud’s enabling technology, virtualization.
VMware declined to discuss their cloud roadmap with me for this story, but as the cliché goes, actions speak louder than words. And VMware’s recent cloud actions are revealing.
You don’t spend a billion dollars to be a bit player. The trouble is, VMware is going up against giants like Amazon and Google who have a head start – a serious head start in Amazon’s case. And there are a slew of other serious contenders, such as Rackspace, Eucalyptus, Microsoft and Citrix (especially after its acquisition of Cloud.com).
So before we look at the roadmap, let’s look at where VMware is now. Virtualization, of course, is the enabling technology that made this cloud revolution feasible. Sure, IT curmudgeons will tell you that the idea of the cloud has been around for years, but it wasn’t practical until virtualization matured.
VMware’s cloud centerpiece is vCloud, which serves as a foundation for building clouds. vCloud competes directly with the Citrix-led CloudStack and with the open source project OpenStack. Muddying the waters, VMware just joined the OpenStack Foundation as a gold member (more on this later).
VMware’s open PaaS solution, Cloud Foundry, shows promise but is still in beta. It will also face stiff competition from established PaaS players, including IBM, Amazon AWS, Red Hat, Salesforce.com, Microsoft and Google. That’s not a list of lightweights.
Where VMware’s cloud potential really starts to shine, though, is with its recent cloud acquisitions, the most important being Nicira, a pioneer of software-defined networking (SDN).
With SDN, the network itself is what is being virtualized. Just as server virtualization decoupled apps from hardware, so too does network virtualization decouple network services and operational control from network hardware. Thus, any available physical network can be transformed into an IP backplane. These virtual networks can then be provisioned to deliver all of the same features of physical networks, but with operational flexibility.
This isn’t trivial. It opens the door (along with virtualized compute and virtual storage) to software-defined data centers.
Nicira isn’t VMware’s only cloud acquisition, but it is the most important. Other cloud-related acquisitions include DynamicOps (cloud automation and management), Cetas Software (Big Data analytics), B-Hive Networks (application performance management), Wanova (virtual desktop management), and Shavlik Technologies (cloud-based IT management). VMware also purchased Log Insight, a cloud analytics platform, from data management vendor Pattern Insight.
What’s the common denominator amongst these acquisitions, other than that in one way or another they all relate to the cloud? The common denominator is management.
And that’s where trouble starts brewing. “Each new acquisition fragments VMware’s cloud management capabilities,” said Shmuel Kliger, CTO of VMTurbo. “VMware is repeating the mistakes made by traditional IT systems management vendors.”
The difference is that when HP, BMC and IBM cobbled together their solutions for managing the non-virtualized data center from an array of point products, data silos were the norm. Enterprises could handle these point products because they already had different storage, networking and server teams, and while everyone paid lip service to breaking down silos, nobody saw this as a huge impediment.
In a software-defined data center, management and data silos pose real problems and seriously undermine efficiency. Kliger isn’t a disinterested party here. VMTurbo provides cloud operations management tools, and, of course, it’s in their best interest to alert people to any cracks in VMware’s cloud management foundation. Nonetheless, his argument is a valid one.
“Instead of solving the management problem, they created a management nightmare. Moreover, most of these tools focus on collecting data – a lot of detailed data – that could be used to alert administrators and generate reports,” Kliger said. “Not only does the collected data not solve anything, but it creates a whole new problem: a Big Data problem.”
With all this data piling up, it must be stored, managed and somehow assessed. With more tools collecting more information, it becomes difficult to extract meaning from the Big Data avalanche.
Another issue with VMware’s cloud strategy is that it recently changed its pricing model. VRAM pricing has been an issue for some time.
Last year, VMware rolled out a new pricing model that charged customers based on the amount of virtual infrastructure, not physical infrastructure, they used. Thus, customers who succeeded in consolidating virtual machines (VMs) onto a few servers were penalized.
At VMworld this year, after vocal complaints from customers, VMware announced that it was doing away with this. It will now charge the old-fashioned way: per-CPU and per-socket.
However, I spoke with an executive at a service provider who complained that the new pricing is not being extended to the VMware Service Provider Program (VSPP).
My source wished to remain anonymous, but he said that now it is service providers, and especially hosting providers, who will be penalized.
“This means that the price service providers give to customers for virtual machines or hosted servers will remain high. The big guys will just move to open source because it’s going to be hard to compete as a VMware service provider paying this memory tax,” he said.
Even the new pricing could be seen as flawed. Eventually, virtualization and the cloud move pricing to a utility model, and the per-CPU model, while probably more profitable for VMware, seems antiquated, a remnant of the shrink-wrap software days.