PayPal’s Internal PaaS Is a Trend, Not an Anomaly

PayPal needs some complex backend infrastructure to support its SaaS operations.

Over the past few months, a number of cloud-service providers—including Google, Amazon, Netflix, and Facebook—have offered up some sneak peaks into their data-center hardware designs, along with a bit of insight into their respective platform decisions.

Paypal is following in their footsteps, highlighting its decision to support its operations with a private Platform-as-a-Service (PaaS) cloud. The revelation came as part of a presentation and audience discussion with Ryan Granard, PayPal of platform engineering, who spoke at GigaOM’s Structure conference in San Francisco.

Whether internal or external, cloud platforms that offer virtualized IT infrastructures have been the rule (or at least the assumption) for most companies. But PaaS has evolved into a more common internal cloud-platform choice recently, according to GigaOM, because they are designed to provide a stable platform on which to run or develop software—and require less effort from IT to build or support.

Companies that go outside for IaaS cloud resources are essentially renting datacenters, while PaaS customers are renting server farms, according most standard definitions of cloud categories. PayPal chose to build an internal PaaS based on Red Hat’s OpenShift software rather than renting external cloud space for development and testing, as many large companies do.

An internal PaaS, separate from most operational infrastructure but not isolated outside the firewall, gives PayPal developers a platform on which to build without the effort of cloning real infrastructure or the security risk of building financial-services apps on virtual servers in someone else’s datacenter, according to Granard.

Red Hat OpenShift comes in external and private versions, designed to let developers focus on code rather than on building or maintaining the platforms on which the code runs. It supports Web Consoles, command-line and IDE development tools, Java, Ruby, Node.js, PHP, Python and Perl for front-end development; it also features a JBoss middleware module to ease connections with existing data or applications, along with a set of optional modules to support apps developed for mobile devices, Web font ends, distributed enterprise designs and big-data analytics.

Having an internal PaaS gives developers instant access to a platform that can realistically mimic a production environment while also acting as a sophisticated sandbox for building software, all without the risk of running it on hardware connected to the operational infrastructure (or the security risk of building financial-services applications on virtual servers hosted in someone else’s datacenter). It also makes life easier for developers, because they can sit down to work on a dev project “and in minutes we have you up and running in a fully connected container” with all the necessary infrastructure already attached, Granard told conference attendees.

Sandbox or not, the apps PayPal developers build on the private PaaS aren’t trivial. Among those built in the sandbox and later released commercially was PayPal Here, a credit-card processing service designed compete with the older, better-established Square.

That doesn’t mean IaaS, even public-cloud infrastructure, isn’t a good option for extra resources to cover spikes in capacity demand, though PayPal doesn’t have a cloud vendor that meets its requirements for security and reliability, he said.

PayPal’s service requirements (and long-term development plans) are such that, rather than just rent capacity from Amazon Web Services or another IaaS platform, the company would probably create a partnership under which it and a cloud provider would customize that latter’s cloud to PayPal’s needs, Granard said.

PayPal is far from the first company to build an internal cloud for application development, according to Forrester Research, which evaluated both IaaS and PaaS cloud services in a cloud-vendor comparison it published earlier this week.

Forrester put both major varieties of cloud platform in the same report because end-user companies evaluating cloud services are looking for the functions they need, rather than boxing products or services into specific categories as vendors and market-researchers tend to do, according to John R. Rhymer, a co-author of the report.

“Cloud platforms don’t fit into neat product categories” Rhymer wrote in a blog posting. “AWS is much more than an IaaS; Microsoft and Google now provide both PaaS and IaaS products. The key to selecting the right cloud platforms for your shop is knowing your developers.”

Developers come in three flavors, the Forrester report added: Coders who want to code without bothering with configuring servers or databases; DevOps pros who can’t operate without configuring the platforms on which their code runs; and Rapid devs—coders who prefer not to code if they can get good graphical or automated tools to do the grunt work for them.

Rather than break cloud services into IaaS, PaaS, SaaS and so on, Forrester classified them according to a developer’s view of the functions that matter, and by differences in the view of developers who value different functions. Rhymer and Forrester cloud-and-virtualization guru James Staten predicted in November that traditional definitions of cloud would become obsolete.

In a furiously competitive, rapidly changing market for IT services, IaaS vendors such as Amazon end up adding PaaS features such as middleware and databases, and end-user customers look for easier ways to integrate public and private clouds. PaaS users demand mix-and-match application services rather than those that come standard as part of a relatively generic services lineup. Hybrid clouds and diversity in the services they offer will become the norm, Rhymer and Staten wrote.

That prediction fits right into the decision of a SaaS provider such as PayPal to build an internal PaaS to supplement its existing infrastructure. It also makes sense for public IaaS or PaaS vendors to add respected datacenter-management tools, including the Chef or Puppet configuration managers, as tools to manage public clouds.

The cloud-morphing won’t stop with acronyms and categories, either, according to a February study from IDC that predicted private clouds would continue to become more popular as a way to manage and make better use of datacenter resources. But not all the private clouds will remain private: the market for private, internal corporate cloud platforms that are hosted with external service providers—some of them also offering multi-tenant public-cloud services—is growing at more than 50 percent per year, IDC found.

 

Image: PayPal

Post a Comment

Your email address will not be published.