The Virtualization Practice | May 23, 2014

The PaaS Build vs. Buy Decision You Didn’t Know You Were Making

It’s no secret that a majority of enterprises are slow to adopt PaaS solutions and favor IaaS solutions instead. What is a well-kept secret is that these enterprises are building their own PaaS solutions on top of IaaS without even realizing it.


Established enterprises have spent years securing data centers, standardizing operational processes, and putting tools in place for things like logging, monitoring, auditing, backup/recovery, and so on. When they build solutions on IaaS for the first time, none of these processes and services exist, and they must be recreated. This is often not obvious to developers until they deploy to production for the first time and then realize that none of these services that they took for granted for years exist in the cloud. To compensate, the developers start building these capabilities into their IaaS cloud environment. Over a period of time, they wind up creating a proprietary platform on top of their IaaS solution.


As these cloud applications mature, more and more PaaS services are added on top of the IaaS solution. Need caching? Install and manage a Redis cluster across multiple availability zones. Need mass email capabilities? Integrate SendGrid with the application. Need to lock down servers from developers to pass audits? Buy Splunk, route all logs to it, lock down all servers, and make Splunk highly available across multiple zones. The list goes on and on. After several sprints of adding “IT plumbing”–type services, enterprises are ending up with custom-made PaaS solutions. A lot of time is spent managing, upgrading, and patching these services while ensuring that they are highly available. Wouldn’t it be easier and more agile to implement an off-the-shelf PaaS instead of building your own?


Take the example above, and see how it would play out if a PaaS were used instead. Implement your favorite PaaS. Need Redis? Call the Redis API and pay for what you use. Need SendGrid? Call the API and pay for what you use. Need Splunk? Call the API and pay for what you use. There is no need to install, patch, and make these services highly available. The PaaS platform does that for you. The time-to-market savings are astronomical. All of the “IT plumbing” sprints go away, and sprints are now dedicated to addressing business requirements.


There are a few reasons. In a post I wrote last year called The Many Faces of PaaS, I declared that there was still much confusion about PaaS, its value proposition, and its maturity. There are a few other issues as well. Enterprises are implementing hybrid cloud solutions, which means many of them require hybrid PaaS solutions. The most mature PaaS solutions are public PaaS players like Heroku, Azure, Engine Yard, etc., which are not meeting the enterprise requirement to run in a multi-cloud hybrid environment. That narrows the playing field down to private and hybrid PaaS solutions like Cloud Foundry, OpenShift, Apprenda, WSO2, etc. The problem here is these enterprises are struggling to get their private clouds up and operational. Every day a company spends trying to get its OpenStack implementation operational is a day it can’t spend implementing a PaaS solution on top of it. I believe that as private cloud implementations become more successful and easier to carry out, enterprises will put more of an emphasis on PaaS.


Many enterprises are building their first generation of cloud solutions. A large percentage of these implementations are built on top of IaaS solutions like AWS, Google, Microsoft, SoftLayer, OpenStack, and others. As these enterprises mature, we will see that many have unintentionally built proprietary PaaS solutions on top of their IaaS vendor’s solutions. Some of these enterprises will enter generation two of their solutions with a lot of lessons learned under their belt, will realize that they no longer want to be responsible for all of this “IT plumbing” work, and will start implementing PaaS solutions. Then we will start seeing more enterprise success stories, driving up adoption and providing clarity to the market of the PaaS value proposition. The only question is when this will happen. I predict we are twelve to eighteen months away from a PaaS gold rush. Many companies are already starting the 2015 budget process or will start in the next few months. The volume of success stories is not yet high enough to drive enough of those budgets.

A great deal of positive progress has occurred recently with various PaaS solutions. Cloud Foundry is getting backed by numerous large corporations. OpenShift is investing heavily in Docker containers. Microsoft is integrating with Apprenda to offer hybrid cloud solutions. These events are driving up proof-of-concept efforts within the enterprise. I expect that late in 2014 we will start seeing more PaaS success stories, which will drive budgets next year. This is all speculation on my part, so take it with a grain of salt. The real point here is that PaaS is coming soon, and when it does, it will be all at once.