Five Takeaways from the JPMorgan Chase PaaS Announcement

jpmorgan chase paasIn one of the more eye-opening recent announcements in the world of enterprise cloud, JPMorgan Chase said last week that it has migrated nearly all of its internal .NET and Java applications to a PaaS model.  Over 430 development teams, 2,000 applications and 4 data centers are now in production in a PaaS  environment delivered by Apprenda.  It is worth noting that rather than relying on a multi-tenant public cloud PaaS service, JPMC has deployed Apprenda on top of existing infrastructure in a private PaaS model.

After digesting the news and commentary, here are five key things we took away from the announcement:

  1. Enterprise PaaS is for real – while many have believed that PaaS models were potentially more interesting from a long-term perspective to enterprises that IaaS, major enterprise deployments have been sparse.  PaaS to date has largely lagged SaaS and IaaS both in enterprise mindshare and adoption.   Rather than private PaaS, most attention has been on public cloud PaaS platforms like, Cloud Foundry and Azure where enterprise use cases have been growing, but limited.   Given this set of adoption dynamics, it was a bit shocking to see a deployment of this size and scope.  While one data point does not a trend make, the fact that JPMC has made this big a commitment to PaaS will likely cause other enterprise CIOs to take a serious look.
  2. There’s big dollars in app dev transformation – as we explored in a recent case study on State Street, standardizing application development on private cloud platforms can drive stunning improvements in cost efficiency and effectiveness.   This appears again to be the case with JPMC, which is using their PaaS environment to drive use of standard application services, architectures and patterns for both .NET and Java applications, which represent about 80% of their portfolio.   The estimated impact so far?  700% improvement in developer productivity and 50 day reduction in time-to-market for new applications.
  3. Private PaaS can drive infrastructure efficiencies – private IaaS and PaaS models aren’t typically seen as big drivers of infrastructure cost reduction.  With no multi-tenant utilization efficiencies, the value of private clouds is typically perceived to be around user self-service and reduction in provisioning times.  That’s why it was a bit surprising to hear JPMC’s private PaaS model has increased infrastructure utilization from 40% to 70%, resulting in 45% reduction in infrastructure costs.  If true, for large enterprises with JPMC’s scale the business case for private vs public cloud may be close if those utilization levels can be achieved and maintained.
  4. Wall Street still isn’t ready for public cloud – none of the JPMC applications migrated will apparently be running in public PaaS environments.  In fact no mention is really made of the fact that Apprenda’s hybrid capabilities can extend private PaaS environment into public PaaS like Microsoft Azure.  There’s no doubt that JPMC’s compliance and performance issues require many, or even most of the applications to remain in a private PaaS environment.  But the fact that apparently none of the 2,000 apps will be run in a hybrid environment raises the question whether private cloud will end up being the end of the journey, instead of a stepping stone to public cloud.  While many large financial services firms seem to be actively engaged in private cloud deployments or pilots, public cloud models still unfortunately seem to be a bridge too far at this point.
  5. It’s all about the developers – instead of a top-down mandate, Apprenda was apparently adopted incrementally on a bottom-up basis over the course of two years.   During this time developer adoption expanded as internal proof points emerged around how PaaS could streamline the application development lifecycle.   It’s interesting to note that usage of the platform has not been mandated by corporate IT, but instead is voluntary.  As a result the internal PaaS platform needs to continue to “compete” against alternative models and approaches.   As successful cloud service providers have discovered, in many cases the key to driving broad adoption is to win the hearts and minds of developers.

Want to learn more about how other large enterprises are using private and public cloud models to drive transformation?  Check out our recent post on how Rio Tinto, CommBank and Ricoh are using cloud to drive enterprise-class value through standardization and simplification.

Want to learn more about how PaaS can drive business and IT transformation in your organization? Contact us at to find out more about our service offerings.



  1. Good observation and conclusion

  2. Wall Street not ready for public cloud? Another possible conclusion is that public cloud isn’t appropriate for wall street.

    Why would a big company rent 1000s of servers 7x24x365?

    • Because they wouldn’t need to pay for them on a 7x24x365 basis. I’m not cheerleading for public cloud here, just saying that it seems like there might be some applications or use cases where it might make sense even for Wall Street…

  3. Marvin Goodman says:

    I might be commenting too late for this to be seen, but I would love to hear some insight – or even guesses – elaborating on this: “That’s why it was a bit surprising to hear JPMC’s private PaaS model has increased infrastructure utilization from 40% to 70%, resulting in 45% reduction in infrastructure costs.” I’m not very knowledgeable about PaaS, but have been around SaaS and IaaS for while. Is the notion here that self-service IaaS, while still offering standardized services, still provides too much subscriber flexibility in what they deploy, resulting in inefficiencies in utilization, whereas PaaS provides them exactly the resources it thinks they need based upon what they’re deploying? Or perhaps its simply a matter of workload placement? PaaS automatically positions workloads for optimal use of shared resources, while IaaS (unless optimized) can inefficiently find various Windows and Linux VMs in the same clusters, inefficiently sharing the same pooled resources?

Speak Your Mind