When Apple built Siri, its intelligent personal assistant application on iOS, it used Apache Mesos the open source cluster manager. Mesos' capabilities allow developers to deploy Siri services in a scalable and highly available manner.
The Mesos project launched in 2009 by graduate students at the University of California at Berkeley. The project was intended to solve two problems: First, the performance limitations that result from static partitioning; Second, the extreme difficulties of fail-over.
Since then, building distributed systems with the project has become essential to today's Application Centric Environment. Mesos works with VMware, Linux, and Microsoft. It's the concept of building a framework and the notion of coordinating and managing clusters as a distributed system in a context.
In the classical model, there is a platform-as-a-service (PaaS) -- examples include Heroku, Google Apps Engine, OpenShift, and Cloud Foundry or infrastructure-as-a-service (IaaS). Mesos sits in the middle of this model. As well as the open container LXC containers which in some ways equated to cartridges.
The key to all these components is orchestration and scheduling. Mesos uses the Apache models, and the industry is also adopting the Kubernetes model. The Mesos model uses the Linux kernel functionality to isolate the compute, memory filesystems and networking functions. Mesos capability is to abstract further the underlying operating system and hardware to allow the creation of a network topology for clusters rather than the parent-child relationship.
Mesos also links easily to Hadoop, databases, and interrelated containers. We are building out a stack that has a management layer sitting on the IaaS and between the PaaS that manages their underlying technology to accomplish the goals established by the Meso project at its inception. It takes away the complexity of running into related services and starts building the model for the application-centric cloud.
The concept of using any of these technologies together becomes complex. Vendors like Red Hat have made efforts to package this technology, and recently announced that it will bundle RHEL 7.0, OpenShift and management together. The directory services and this stack require integrated components to manage and distribute the containers in the Mesos clusters. Kubernetes or other technologies can be used for orchestration. The Mesos model requires a dedicated machine to not only orchestrate, but to manage the groups, the directory, the workflow and the integration into the PaaS platform, containers, and the IaaS platform.
This architecture lets you build at the scale of Amazon Web Services. It also allows you to deploy the clusters on different environments provided connectivity so that each container can run according to its needs. Each environment has a different abstraction connected in a virtual service by way of a service bus connection.
The new model requires coordination by the industry than we've seen in the past. The joint work is essential for creating this operating environment and removing complexity. Just as with containers we are going to see convergence between OpenStack, Azure, VMware, Amazon and Google all supporting the tools and the elements of a platform-as-a-service. The application will become the focal point of the cloud, not the underlying infrastructure. It's the application that drives whether or not you use a container approach, a virtual machine approach, a native operating system approach or a hybrid approach to run these applications. It is very clear after the launch of the Open Container Project that collaboration is beginning. We must, as an industry understand that there is no one vendor that can tackle this problem and produce an application-centric cloud without collaboration.
The APIs that drive the cloud in the future need to be normalized, and the Internet of Things is 100 percent reliant on this effort. In the future, I will explain how the Internet of Things fits into an application-centric cloud. The managed service providers need to adopt this model to thrive in the new cloud era.