Cisco Application Enablement Platform as a Service
Every company is becoming a software company. The Intercloud’s primary use case is to enable application portability and interoperability across a world of clouds. By virtue of its broad definition, one can throw in various solutions under the PaaS umbrella. For our purposes we will define what PaaS encompasses by looking at what customer problems need to be solved and working backward from there. Our target audience in terms of usage of the PaaS solution on the Intercloud is the enterprise developers and enterprise IT engineering teams on the customer side; the set of decision makers for adoption of a particular PaaS offering will have a mix of technical and business roles. Figure 2-8 shows the traditional differences between infrastructure and PaaS.
Figure 2-8 IaaS and PaaS Differences
We will start by looking at what problems the first group faces and then move on to selecting the PaaS components that will solve those problems. The following three phases clearly distinguish between PaaS and IaaS and how Cisco Intercloud overcomes many of the issues:
Phase 1 is the development phase, which is when the developers and testers develop, build, deploy, and test the application in their local environment or a preproduction version of the cloud environment they plan to use. One friction point here is that they need to be able to consume the various cloud components from their application code, but doing so involves repeatedly writing code that should ideally be abstracted away in a utility layer. Another pain point at this stage is that the developers need to have confidence that their application that worked correctly in the test environment will continue to do so when deployed to the production environment. The Intercloud addresses these problems by packaging OpenStack APIs (the framework under which the Intercloud is standardized) that let an application request, consume, and release Intercloud resources in software development kits (SDKs) built specifically for the common languages used by customers (for example, Java or Python). The second problem of ensuring consistency through the various deployment stages is best solved by providing a facility for customers to containerize their application so that it deals with the underlying infrastructure only through interfaces previously agreed on between the container and the infrastructure (in this case OpenStack).
Phase 2 (deployment phase) is the deployment of the built and tested application through various stages, culminating in the production cloud environment. This is similar to the previous phase where developers require consistency in the behavior of their application but in addition are interested in having a deployment experience that is not cumbersome, error prone, and manual in nature. An ideal solution here will provision the required cloud resources, wire them up as needed, deploy the right containers, and provide the customer a report that lists the deployment status, the resources being used, and the endpoints to the applications. All of these would be driven via preconfigured settings to ensure a high level of automation.
Phase 3 (support phase) starts after the deployment is completed and the application is supporting production traffic (or workloads for an offline application). During this phase it is important to ensure that cloud resources are scaling in an elastic manner, key health parameters are monitored and alarmed upon, and the engineering team is able to quickly fix issues by deploying patches or new versions with minimal downtime. The scalability aspect can be handled by leveraging the autoscaling functionality in OpenStack and ensuring that it works with the container format under which the customer has chosen to deploy. When it comes to monitoring and triggering alarms, the customer should be able to define metrics around system resources (such as CPU usage) and application-specific measurements (such as average latency when communicating with another node). This requirement means there must be a monitoring service that customers can easily integrate into their application and can be viewed in multiple ways (for example, a dashboard for a high-level view, logs for deep diving). Last, the chosen container format should be able to deploy updated containers to a production environment with zero downtime to the service. This monitoring is key to application intent enforcement and improved performance of the application.
Integrated Platform as a Service (iPaaS)
Integrated platform as a service (iPaaS) is defined as components of cloud services enabling development, execution, and governance of integration flows connecting any combination of on-premises and cloud-based processes, services, applications, and data within individual or across multiple organizations.
The primary use case here is to integrate the enterprise PaaS, public PaaS, and application deployments into the development lifecycle. The Intercloud will enable integration between select partner offerings Cisco wants to expose to its customers as “first-class” PaaS offerings and will deeply integrate them into the Cisco ecosystem, including giving them a presence in the Horizon dashboard. An example of this would be running the Red Hat PaaS product OpenShift service in a hosted model that doesn’t require customers to have a dedicated OpenShift controller and will let them choose from a list of “cartridges” that Cisco maintains in its library.
In this same scenario, if customers prefer a single-tenant PaaS controller, they will always have the option to buy OpenShift from the Cisco Intercloud Marketplace for their exclusive use. Another example would be Cisco maintaining its own “hub” and container repositories for the Docker container format for customers to build and manage their applications. By pursuing this option, Cisco also has the ability to integrate the PaaS solutions to provide a seamless experience for its customers (for example, the Cisco hosted OpenShift can easily fetch containers that a customer owns). When Cisco sees an opportunity to differentiate (such as a unified dashboard for all “first-class” PaaS offerings), it is able to build that and market it exclusively to its customers. It is also worth noting that if a partner’s solution is not exposed as a deeply integrated solution on Cisco Cloud Services, the partner is still able to reach the Cisco customers by onboarding on the Cisco Marketplace; this provides Cisco customers with options that suit their workload and interoperability needs.
The availability of competing PaaS solutions in the Cisco Marketplace will also help Cisco maintain its leverage with the vendors it has chosen to supply its “first-class” PaaS offerings. The downside of this option is that it requires Cisco to get buy-in, at a preferred price point, from the partners it wishes to integrate. This can also lead to Cisco having to build and operate parts of the PaaS when it simply cannot find a good external offering.
Application Platform as a Service (aPaaS)
Application platform as a service (aPaaS) is defined as a category of cloud computing services that provides a platform allowing customers to develop, run, and manage applications without the complexity of building and maintaining the infrastructure typically associated with developing and launching an application.
The primary use case here is derived by understanding that the Intercloud is trying to become the cloud-of-clouds platform and the IoE platform. To enable this platform, a curated experience with all the PaaS capabilities will need to be created. Cisco will create a container format that works on all OpenStack environments in a consistent manner and will provide developers a set of SDKs to build against and package toward this container. Cisco will then augment the capabilities of OpenStack’s orchestrator, Heat, to deploy these containers to cloud. Next, Cisco will build and operate a service that works in conjunction with Nova (OpenStack’s compute module) to ensure autoscaling of the underlying resources. Finally, Cisco will have a homegrown monitoring service that will look for deviation from preconfigured steady states in addition to executing automated healing steps. This is similar to what Amazon Web Services (AWS) has in the form of Elastic Beanstalk, EC2 Auto Scaling, CloudFormation, and CloudWatch.
The benefits of this option are that Cisco can fully control the customer experience, iterate fast on bug fixes and new features, and work with otherwise incompatible systems (for instance, integrate with Cisco NFV modules). Furthermore, by virtue of offering innovative features specifically for its enterprise customers, this option also enables Cisco to differentiate from other cloud offerings more prominently. The downside here is that this will take a larger monetary and time investment to bring to market. Furthermore, this will make it harder for Cisco customers to realize the Intercloud benefits by making it difficult to deploy their applications to other clouds, both private and partner, unless Cisco makes a concerted effort to ensure that its PaaS is available in those clouds.
Use Cases for PaaS
The following use cases for PaaS are covered by the Intercloud:
Cloud native: Most developers are looking at the role of cloud-native development in the next generation of enterprise applications and mobile devices. This model does not mean that all of your development is done in public clouds; it is more about how a developer approaches code development, deployment, and runtime. Containers are becoming very critical to cloud-native developers because of the abstraction layer they provide, and the individual services greatly simplify development. This use case is more likely to use an aPaaS framework as it is more aligned with developers writing to application platforms rather than the underlying infrastructure.
Cloud valid: Several enterprises are on a journey to cloud by enabling many web application and public-facing services. This use case has more to do with the legacy applications in the enterprise and determining which applications should be exposed in clouds and which ones cannot, at least not yet. The primary reasons for not having those applications in clouds mostly have to do with the internal OSS/BSS dependencies, database dependencies, or data security concerns. The Intercloud enables this use case by iPaaS, which enables enterprise customers to leverage the infrastructure APIs in the enterprise and in cloud to create applications that can be deployed behind the firewall or in clouds (public or private), thereby creating a hybrid cloud use case where “hybrid” is defined as application components deployed in the enterprise data center and cloud connected securely.
Hybrid DevOps: Hybrid DevOps is the Intercloud IoE platform for enterprises to start developing today for internal and external deployment patterns that over time will move more and more toward cloud native or the next-generation development pattern (fourth generation of PaaS). DevOps is an enterprise software development phrase meaning a type of agile relationship between development and IT operations. The capabilities of the Intercloud for the hybrid DevOps use case are displayed in Figure 2-9.
Figure 2-9 Intercloud Hybrid DevOps Capabilities