IBM, BEA (now part of Oracle) and RedHat have done an amazing job at achieving strong penetration in the Java app server market, having spent 15+ years building out those businesses. In addition, open source projects like Tomcat have achieved unparalleled market share. What we’re observing with this and RedHat, who recently made OpenShift generally available, is that PaaS is a natural extension of the app server into cloud. Does being successful as an app server vendor make the success of Java PaaS a default outcome? No. In fact, it will likely be detrimental to their success in the enterprise PaaS market.
First, let’s remember one major failure of the app servers of yesteryear: fragmentation. Java promised a “write once, run anywhere” outcome. In reality, we ended up with an outcome closer to “write once, run only here” or “write once, debug everywhere.” If you wrote an app targeting a specific app server, you were bound to that app server. The vision of Java EE didn’t quite produce the compatibility we hoped for. This lack of homogeneity in the Java market is the bane of Java in enterprise IT. Different teams and lines of business use different app servers, generally, and attempts to standardize on one (such as WebSphere), haven’t panned out in practice. What you end up with is an enterprise IT function that is absorbing the cost of this fragmentation.
One of PaaS’ general promises is to normalize workflows, infrastructure patterns and architecture patterns across a portfolio of applications in an enterprise. Enterprises are looking to private PaaS to 1) abstract away infrastructure details and 2) homogenize the application portfolio as much as possible from a development and management standpoint. Enterprises have hundreds, if not thousands, of legacy apps targeting these app servers. Unfortunately, legacy app server vendors like RedHat can’t help with this. Under the hood, RedHat will tailor their PaaS to either exclusively support – or best support – their app server JBoss. As TechCrunch’s Scott Merrill points out here:
”Red Hat has their own Java application server in JBoss, and OpenShift Enterprise is certainly tailored to make this a well-behaved component. Users of IBM WebSphere or Oracle’s WebLogic will have greater challenges adopting OpenShift for their needs.”
The approach of best supporting their own app server will be the approach that each of these big app server vendors take. Outcome? Fragmentation in the app server market will rear its ugly head in the PaaS market. Yes, proponents of tech like OpenShift will claim that their “cartridge” model will allow you to plugin something like IBM WebSphere, but what happens when it doesn’t work? Can you call IBM? Will RedHat provide customer support for WebSphere on OpenShift? Are they even equipped to do so? Doubt it. The fact is, these traditional app server vendors are slaves to their legacy and their revenue, and will build enterprise PaaS to bolster that legacy rather than shape their future. For RedHat, PaaS is just a vehicle to sell more RHEL.
If an enterprise buys into a PaaS from a legacy app server vendor, they will then have to buy into some sort of IBM PaaS, and an Oracle PaaS, and one that’s Tomcat-based to optimize for those workloads. Running a private PaaS mapped to each app server makes little to no sense, since it would perpetuate the fragmentation and defeat one of the core goals of deploying a PaaS based private cloud. Additionally, add that each enterprise would then need another PaaS for .NET on Windows, and the mess gets…well, messier. Because of rivalries with Microsoft, and lack of competency in Windows and .NET, it’s unlikely that you’ll see any of the legacy Java app server vendors supporting native .NET or SQL Server deployments in anything more than a superficial way.
Ultimately, this means that independent PaaS vendors like Apprenda have a tremendous structural advantage: we can build, and properly support, a PaaS that provides heterogeneous app server support coupled with cross platform support for Windows and .NET. According to Gartner, “Microsoft’s and Oracle’s Java software platforms are practically de facto standards in at least 85% of large-scale — Global 2000-class — IT organizations…Through 2017, the Microsoft and Java software platforms will continue to enjoy at least 85% adoption among Global 2000 IT organizations.” Ultimately, this means that an independent PaaS can plug into an enterprises’ application portfolio and developer community much more cleanly and broadly, leading to a drastic difference in ROI potential.