Pimp Your PaaS — The Race to Something Different

September 6, 2010 Off By David
Object Storage
Grazed from GigaOM.  Author: Gary Orenstein.

Now that VMworld has come and gone, we can rest assured that there are plenty of choices for cloud infrastructure build-out, including VMware’s vCloud Director. But the cloud computing land grab has a more nuanced and intriguing battle underway–the race to deliver differentiated Platform-as-a-Service (PaaS) offerings to woo developers and build critical mass as quickly as possible.

While there is — and will continue to be — a significant amount of spending to tinker with and repackage enterprise applications into virtual machines for cloud migration, I believe the long-term stickiness the major vendors seek is a compelling PaaS offering. This is just part of the reason VMware has aggressively pursued acquisitions to beef up vFabric, including SpringSource for Java application development, GemFire/GemStone for data management, and RabbitMQ for messaging and communications between application nodes.

Much in the way that Microsoft dominated the business applications space for decades, the new race for the next decade or two is capturing the hearts and minds of application developers and their efforts on new cloud deployments. We’ve seen the bulking up of PaaS offerings across the board. Salesforce.com’s Force.com boasts a broad ecosystem of partner applications and solutions. Google’s AppEngine has been embellished with services such as Memcache, the popular caching tool, and a Mapper API for running MapReduce operations. Amazon continues to add new services on top of its popular EC2 compute platform including database services, MapReduce, and messaging. There is also, understandably, a massive ecosystem around the suite of Amazon APIs.

According to Adrian Cole, founder of Jclouds, an open-source framework of APIs for portable abstractions of cloud-specific features, “The most important thing PaaS providers can do is keep application developers in their zone. The last thing they want to happen is for a development team to be delayed for months while they sort out a low-level infrastructure issue.”

In pursuing this goal, we will likely see major PaaS players continue to bulk up their offerings through new features or acquisitions. One language isn’t likely to cut it. Google’s AppEngine added Java support last year in addition to its original choice of Python, and Microsoft touts developer choice across .NET, PHP, Ruby, Python or Java. “To give developers and companies the reassurance that they are not going down a one-way street, the multi-OS or multi-language PaaS must become a reality,” continues Cole. “And while developers may have been attracted by carrots before, they are now all foodies and require much more. That’s why we’re likely to see more innovation around and expansion of PaaS offerings in the future.”

The dark horse in this race might be Oracle, and it probably has time. PaaS initiatives need to get into the corporate budget, through the experimentation cycle, and into production which could take time for the industry at large. This race is far from over, and we may not have even seen a top contender emerge.