From its beginning, Cloud Foundry has been committed to providing developers and enterprises choice of deployment options spanning public and private clouds. In this post, Jonas Partner, CEO of CloudCredo and OpenCredo, shares how their enterprise customers are using PaaS to deploy applications that span public and private clouds to increase availability, add extra capacity and prevent lock-in.
As one of the co-founders of OpenCredo I have been actively working with Cloud Foundry since its early days. We were one of the first companies to launch a production application on Cloud Foundry with the help of our friends at Carrenza. More recently we have established CloudCredo to help both enterprise customers wishing to host Cloud Foundry on their own infrastructure and cloud service providers wishing to offer Platform as a Service. We helped Ospero to bring a Cloud Foundry-based PaaS running on vCloud Director to market to better serve their customers.
With the growing availability of public Cloud Foundry instances, we are seeing an increase in interest from enterprise customers in hybrid cloud as a way to extract maximum value from their existing infrastructure while enabling capacity on demand through expansion into public cloud. In this blog we will discuss why hybrid PaaS with Cloud Foundry is gaining interest, some of the ways people are using it today and how we hope to be using it in the future.
Why Cloud Foundry for Hybrid PaaS
The availability of a PaaS that seamlessly spans in-house data centers and various public cloud offerings makes hybrid cloud a compelling commercial proposition. For example, it serves to drive competition among public cloud providers. Extra capacity can be bought from the most cost competitive provider, or spread across a number of providers, as determined by the best commercial fit for the need. To date the majority of public PaaS offerings have been closed source and can’t be run within the enterprise data center. Combine this form of lock-in with some well-publicized outages and most large enterprises simply aren’t willing to consider betting everything on a single instance cloud provider. We see the open, extensible nature of Cloud Foundry and the increasing number of companies offering hosted Cloud Foundry instances as key to making enterprises comfortable with PaaS.
For us the number of companies offering or intending to offer a Cloud Foundry instance is reaching a critical mass, giving enterprises a competitive set of choices of where to run their applications. The Cloud Foundry Core initiative also means that customers can rely on public Cloud Foundry Core compatible instances to meet certain minimum requirements. Taken alongside improved support for running Cloud Foundry on public IaaS such as EC2, enterprises now have three real options for deploying cloud applications: 1) on a Cloud Foundry instance running in house on existing infrastructure, 2) on a Cloud Foundry instance running on large scale public IaaS such as EC2 or 3) on one of the increasing number of cloud providers offering Cloud Foundry as a service.
Being able to pick a company offering a PaaS hosted within national boundaries is already broadening the applicability of PaaS to regulated industries that have requirements on data sovereignty and how data is managed. There are still many other considerations around technology stacks and non-functional requirements that determine whether Cloud Foundry is suitable for a particular workload.
Simple Hybrid Cloud Use Cases
We are seeing significant interest in hybrid PaaS as a way of adding capacity in peak periods and increasing availability. Currently this only works out of the box for simpler cases where 1) there is no need to massively scale the persistence technology and 2) the application’s database does not need to be shared between services on either side of the public private cloud divide. Where enterprises employ SOA it can also be challenging to find a subset of the services that can be moved out to the public cloud. While services can span public and private cloud, this adds complexity, potential performance challenges and cost where data stored in the public cloud is separately billable.
The factors that determine the simplicity of an application for the purposes of a hybrid cloud scenario are primarily how coupled it is to the rest of the enterprise. This coupling can take the form of persisted data, messaging and application services. The simplest case is an application that has none of these forms of dependencies – with the increasing adoption of SOA it is increasingly rare to find these types of applications. When considering the potential use of hybrid cloud, the cost benefit versus the complexity due to connectedness to the rest of the enterprise will be one of the key trade-offs.
Unless for some reason you have a large number of island-like applications then batch processing applications can often provide a fertile ground in the early days of adopting a hybrid cloud deployment strategy. Batch applications often combine minimum data connectivity to the rest of the application, often running on a data snapshot, with providing cost benefit from scaling up just for the batch run.
Scalable Database Services
The open source Cloud Foundry project provides a number of open data technologies that can be configured to “scale up” as single instance services out of the box. We see incorporation of scale-out, multi-instance database services as key to widening the use cases that can be addressed by Cloud Foundry and are currently working to address this opportunity for our customers.
At CloudCredo we recently open sourced a simple Cassandra service integration for Cloud Foundry on GitHub. The initial implementation was targeted mainly at development and test usage since it was single instance and therefore didn’t provide any way to make the database scale out or highly available. We are now extending that out to a multi-node cluster and we intend that other services such as MongoDB will also follow.
Scalable Databases as a Service can be plugged in to Cloud Foundry very easily by leveraging the Service Broker. This pattern allows the cluster to be managed from outside of Cloud Foundry but for the ‘external’ cluster service to be exposed to clients as a running Cloud Foundry service. The benefits of this model include Cloud Foundry features such as central configuration management and automatic service binding from applications.
While horizontal scaling and clustering of open database services opens up new potential, what really excites us is the potential to use the multi-data center capabilities of persistence technologies in conjunction with hybrid PaaS. Cassandra and other open data technologies such as Riak make it possible to have a logical database which spans multiple data centers and therefore potentially both public and private cloud.
A PaaS which integrates multi-data center databases that are easy to deploy and provide data synchronization would make solving problems such as disaster recovery and transparent scaling of both database and application code much easier. Disaster recovery currently relies on expensive low latency RPO systems that require manual, offsite back-ups be taken to mirror production. A multi-node, multi-data center data service would have the potential to automatically mirror data across data centers as part of the online transaction processing of the application. Having effectively the same database available within both public and private PaaS would also simplify the process of scaling out.
Conclusion
The growing popularity and increasing number of ways to consume Cloud Foundry is making hybrid PaaS an appealing reality for many enterprises. Cloud Foundry provides the Service Broker to facilitate integration of new, clustered open data technologies and legacy clustered databases, easing enterprise adoption in cases where the database services that come as part of the platform are not sufficient. The combination of reliable open source technologies that can deal with databases and the increasing availability of Cloud Foundry is an exciting option for those trying to use PaaS within the enterprise. Being able to solve these sorts of traditionally expensive problems with a PaaS will make PaaS even more compelling for large enterprises.
About the Author: Jonas Partner is CEO of OpenCredo, a software consultancy and delivery firm with offices in the United Kingdom and North America. Recently, Jonas co-founded CloudCredo, a company focused on helping businesses get Cloud Foundry up and running on the infrastructure of their choice. Before establishing OpenCredo in 2009, Jonas was a consultant with Spring Source where he contributed to the Spring Integration project. Jonas maintains interests in Machine Learning and the scalability of highly concurrent enterprise systems.
