Why the PaaS market as we know it will not die off
I’ve been hearing a lot about Platform as a Service (PaaS) lately as part of the broader discussion of cloud computing from both customers and in articles across the web. In this post, I’ll describe PaaS, discuss a recent article that came out on the subject, and take a shot at sorting out IaaS, PaaS, and SaaS.
What is PaaS?
First a quick trip down memory lane for me. As an intern in college, one of my tours of duty was through the manufacturing systems department at an automaker. I came to work the first day to find a modest desktop computer loaded with all of the applications I needed to look busy, and a nicely printed sheet with logins to various development systems.
My supervisor called the play: “I tell you what I want, you code it up, I’ll take a look at it, and move it to test if it smells okay.”
I and ten other ambitious interns were more than happy to spend the summer with what the HR guy called “javaweb.” The next three months went something like this:
Part I: Set up the environment…
- SSH to abcweb01dev.company.com, head over to /opt/httpd/conf/httpd.conf, configure AJP to point to the abcapp01 and 02dev.company.com
- SSH to abcapp01.dev.company.com, reinstall the Java SDK to the right version, install the proper database JARs, open /opt/tomcat/conf/context.xml with the JDBC connection pool
- SSH to abcdb01dev.company.com, create a user and rights for the app server to talk to the web server
- Write something simple to test everything out
- Debug the environment to make sure everything works
Part II: THEN start coding…
- SSH to abcweb01dev.company.com, head over to /var/www/html and work on my HTML login page for starters, other things down the road
- SSH to devapp01dev.company.com, head over to /opt/tomcat/webapps/jpdwebapp/servlet, and code up my Java servlet to process my logins
- Open another window, login to abcweb01dev and tail –f /var/www/access_log to see new connections being made to the web server
- Open another window, login to abcapp01dev and tail –f /opt/tomcat/logs/catalina.out to see debug output from my servlet
- Open another window, login to abcdevapp01 and just keep /opt/tomcat/conf/context.xml open
- Open another window, login to abcdevapp01 and /opt/tomcat/bin/shutdown.sh; sleep 5; /opt/tomcat/bin/startup.sh (every time I make a change to the servlet)
(Host names and directory names have been changed to protect the innocent)
Setting up the environment was a little frustrating. And I knew that there was more to the story; some basic work, call it Part 0, to get some equipment in the data centre, the OS installed, and IP addresses assigned.
Part I, setting up the environment, is the work you would do to setup a PaaS platform. As a developer, the work in Part I was to enable me and my department to do the job in Part II – and we had a job to do – to get information to the guys in the plants who were actually manufacturing product!
So, here’s a rundown:
Part 0: servers, operating systems, patches, IPs… IaaS
Part I: middleware, configuration, basic testing… PaaS
Part II: application development
So, to me, PaaS is all about using the bits and pieces provided by IaaS, configuring them in a usable platform, delivering that platform to a developer so that they can deliver software to the business. And, hopefully the business is better off because of our software.
In this case, our software helped the assembly plant identify and reduce “in-system damage” to vehicles – damage to vehicles that happens as a result of the manufacturing process.
Is the PaaS market as we know it dead?
I’ve read articles predicting the demise of PaaS altogether and others just asking the question about its future. There was a recent Networkworld article entitled “Is the PaaS market as we know it dying?” that discussed the subject. The article makes three main points, referring to 451 Research, Gartner, and other sources.
- PaaS features are being swallowed up by IaaS providers
- The PaaS market has settled down while the IaaS and SaaS markets have exploded
- Pure-play PaaS providers may be squeezed from the market by IaaS and SaaS
I agree with point #1. The evidence is in Amazon Web Services features like autoscaling, RDS, SQS, etc. These are fantastic features but interfacing to them locks developers in to using AWS as their single IaaS provider.
The IaaS market is still very active, and I think there is a lot to come even though AWS is ahead of other providers at this point. IaaS is commodity, and embedding specialized (read: PaaS) features in an otherwise IaaS system is a tool to get customers to stick around.
I disagree with point #2. The PaaS market has not settled down – it hasn’t even started yet! The spotlight has been on IaaS and SaaS because these things are relatively simple to understand, considering the recent boom in server virtualization.
SaaS also used to be known as something that was provided by ASPs (Application Service Providers), so many people are already familiar with this. I think PaaS and the concepts are still finding their place.
Also disagree with point #3, the time and opportunity for pure-play PaaS providers is now. IaaS is becoming sorted out, and it is clearly a commodity item. As we highlighted earlier, solutions from PaaS providers can ride on top of IaaS. I think that PaaS will be the key to application portability amongst different IaaS providers – kind of like Java: write once, run on any JVM (kind of). As you might know, portability is one of NIST’s key characteristics of cloud computing.
Portability is key. I think PaaS will remain its own concept apart from IaaS and SaaS and that we’ll see some emergence of PaaS in 2014.
Why? PaaS is the key to portable applications — once written to a PaaS platform, it can be deployed on different IaaS platforms. It’s also important to note that AWS is almost always associated with IaaS, but they have started to look a lot like a PaaS provider (I touched on this in a blog earlier this month). An application written to use AWS features like AutoScaling is great, but not very portable.
Lastly, the PaaS market is ripe for innovation. Barriers to entry are low as is required startup capital (there is no need to build a datacenter to build a useful PaaS platform).
This is just my opinion on PaaS — I think the next few years will see a growing interest in PaaS, possibly even over IaaS. I’m interested in hearing what you think about PaaS, feel free to leave me a comment here, find me on twitter at @dixonjp90, or reach out to us at firstname.lastname@example.org