Cloud Session: Open Source in the Cloud
Slides of the presentations are available on the main program page.
Schedule: Wednesday November 24, 2010 - 01:30pm - 06:00pm.
Session Chair: Yin Gang, NUDT (National University of Defense Technology).
1. Cloud aware large scale distributed SOA
Schedule:01:30pm - 02:00pm
Speaker: Christophe Hamerling (R&D Engineer & Project Leader, PetalsLink).
Abstract: The Internet is growing fast and is becoming the support for thousands to billions of services prosumers. In parallel, business entities needs to collaborate more, are using Internet as a communication layer and are starting cloud intiatives to expose and consume IT services into public and private clouds.
The goal of this presentation/talk is to introduce state of the art solution and to propose OW2-based Open Source response allowing entities to develop and extend their business easily in a cloud-aware way. The approach we are proposing is composed of three major modules : A cloud-aware/federated service bus, a service governance tool and an online Collborative Business Process Editor.
The cloud service bus extends the OW2-Petals ESB by enabling service exposition and consumption between service farms by using OW2-ProActive as the framework to build a federated communication layer. Once communication links are established between parties using the cloud middleware, services visibility can be expressed using the OW2-Petals Master governance solution and automatically exposed to defined partners by the service bus. Finally, an online business processes editor is deployed in the cloud and fully connected to the infrastrucutre in order to provide a fully collaborative processes creation using BPMN standards.
2. From Centralized to Distributed Cloud"
Schedule: 02:00pm - 02:30pm
Speaker:Jean-Paul Smets, Free Cloud Alliance
Abstract: tbc
3. ONCE-PaaS: A PaaS Solution Stack 4 Cloudware.
Schedule:02:30pm - 03:00pm
Speaker: Wei Wang (Institute of Software, Chinese Academy of Sciences (ISCAS)).
Abstract: Many features of cloud computing, including multi-tenancy, easy deployment, pay per use, elasticity…, will bring great benefits to both service user and provider, however, also cause technical challenges for cloudware development and hosting. Several gaps between current cloud computing technology and what is needed including (1) supporting resource isolation and security isolation for high level multi-tenant style cloudware, (2) supporting self-* properties for cloudware development, deployment and management, and (3) programming model support for “cloudified” applications.
The ONCE-PaaS, which is provided by ISCAS, delivers open source cloud computing solutions aiming to solve these problems. It has been developed through several research projects across the cloudware lifecycle:
- It provides concurrency control and access control for tenant isolation. This enables deploying multiple instances of the same application on a single application server instance (one application instance per tenant), and therefore, allows to scale to additional tenants quickly and will be more cost effective and less overhead than the virtualized approach.
- It provides autonomic mechanisms and tools to manage cloudwares through their entire life cycle, from deployment to monitoring, metering, and resource provision. It supporting QoS-driven service selection for designing composite service. It proposes a model-driven approach to assemble cloudware components inside customizable hosting environment, and validate deployment constraints and eliminate deployment conflicts automatically. It provides an elastically scalable and fault-tolerant compute infrastructure which dynamically allocates more hosts and performs the necessary load balancing. This also allows tenants only pay for what they need.
- It provides a cloud-aware programming model supporting distributed concurrency programming. This allows developer to develop distributed cloudware without taking care of the scalability and autonomic capabilities himself.
The ONCE-PaaS solution stack, which has been developed through several research projects across various cloudware hosting layers, is going to collaborate with several OW2 projects(JOnAS, JASMINE) for OW2 Open Source Cloudware Initiative (OSCi).
4. Starting for the Cloud
Schedule:03:00pm - 03:30pm
Speaker: Minghui Zhou, Peking University.
Abstract: Nowadays, cloud computing is recognized as a promising computing paradigm to run web applications. It enables the applications to access the computing resources on demand with little efforts, so that resource over or under provision are avoid and high availability, scalability and adaptability can be achieved by the applications. The magic behind cloud computing is resource sharing: as applications?workloads usually vary all the time, and their variations are irrelative, it is possible to share a large pool of configurable resources among the app lications. However, a Cloud Computing platform is more than a collection of computer resources because it provides a mechanism to manage the resources. As the runtime platform of applications, middleware is required to take on the resource management responsibility.
The goal of our research is to explore how to extend middleware to support cloud computing. In our research context, the cloud can be seen as a shared cluster where a large number of applications are hosted, and each application can run on a subset of nodes and each node can multiple applications simultaneously. Resource management in such cluster includes four steps, which are resource monitoring, resource demands estimation, resource plan and resource allocation. And our work focuses on the resource plan and allocation. Until now , we have proposed a resource management method for the shared cluster, and in this term, we are going to extend JOnAS 5 with this method to support shared cluster.
In the resource management method, first, we propose a self-adaptive resource plan approach which borrows the idea from human market, and have done some simulation experiments to test the effectiveness and the performance of the approach. Now, we are implementing our resource plan approach on JOnAS 5.
Second, we propose to integrate a virtual machine monitor, named OpenVZ with JOnAS 5 in order to allocate each node ? resources to its hosted applications at runtime. Each virtual machine runs an application and a customized JOnAS. Upon receiving the resource plan at runtime, JOnAS invokes the interfaces of virtual machine monitor to allocate the resource accordingly.
Third, we are going to utilize the on demand support provided by JOnAS 5 to dynamically customize the middleware for the applications, because sometimes the cluster needs to adjust its resource allocation through changing the placement of applications on the nodes, and before deploying the application on a new node, the middleware needs to deploying the middleware services by the applications first.
Break: 03:30pm - 04:00pm
5.A Cloud Platform for Delivering Instant Development Service with Service Oriented Approaches.
Schedule:04:00pm - 04:30pm
Speaker:Hailong Sun, BeiHang University.
Abstract: Cloud computing is commonly characterized as a three-layer architecture including IaaS, PaaS and SaaS, while service oriented approach is widely considered as a promising software development method. In this talk, we report our experience of moving traditional service oriented software development to the cloud computing environment. Specifically, we are working on building a cloud platform for service oriented software development. This platform can provide software developers with a development environment on demand, which will greatly facilitate development work and improve the software productivity. Our primary goal is to provide instant development, instant deployment and instant running services for service oriented software developers. Corresponding to the three layers of cloud computing, our work includes software appliance management, app engine and online development environment. We elaborate on the design and implementation experience.
6. XWiki: a perfect match for collaboration in the Open Cloud
Schedule:04:30pm - 05:00pm
Speaker: Ludovic Dubost, CEO, XWiki.
Abstract: Moving applications to the cloud creates many challenges for users: Does the application fit my need? Is it easy to use? Can I customize it? Are my data secure? How can I move away from my cloud provider with my data and still use them efficiently?
These questions are particularly important in the Collaboration Space and when working on Business Critical Information. All collaboration applications need to be closely adapted to the company's processes. Once your collaboration tool grows in usage, you might need to take back control or change cloud provider.
XWiki's collaboration solution is a perfect match for the Open Cloud. This session proposes 1) to go through the key aspects of open cloud collaboration, 2) to focus especially on the development features of XWiki which allows extensibility right in the cloud, 3) to explore the ways developers can directly online participate in the development of the solution (by creating and integrating applications) and for users to access easily to an open and innovative space.
7. Cloud Infrastructure SetUp, the start of making Money
Schedule:05:00pm - 05:30pm
Speaker: Jose Luis Serano Fernandez (Consultant).
Abstract: OpenSource has became a Flag that all Players are flying. Which are all these players and which are the new relationships created around the "Cloud". How the global market has become local when you need to make money? Why are the small players so important and why the Cloud, at last, depends on local small players?
Introduction of a Spanish Company Project with the marketing and technological aspects that give the opportunity to reach market. How procbel (http://www.procbel.com) has faced the deal to "Cloud" the software products to suit business needs.
The message to customers, the message to Partners. The main aspects of the pricing and control issues.
About infrastructure How to integrate the Hardware, Communications and Software, (Open Projects and vendors one) How suits the main Software infrastructure Eclipse RCP, Spagic, SpagoBI and openQRM.
The new procedures the company should implement, (selling, provisioning, ISO, and so on).
8. NEXUS: To integrate cloud and grid
Schedule:05:30pm - 06:00pm
Speakers: Kailash Selvaraj, Tech. Staff, Madhusudhana Rao Sriramagiri, Project coordinator, CDAC, India.
Abstract: The proposed NEXUS middleware component acts as an interface to the cloud and grid middleware. The design adapts to the cloud SLA and its credentials, and to grid security mechanism and scheduling policies.
The functionalities of the middleware component includes from the cloud perspective elasticity, distributed caching, basic and advanced scheduling, instances and its network management, accessibility, workflow engine, service provisioning, and from the grid perspective virtualizes the grid services, agent to interact with the grid schedulers.
This component is responsible for virtual organization formation in a hybrid environment, advanced reservation, event logging, and dynamic job migration.
The computational jobs run in a grid environment under the management of its grid middleware, local and Meta scheduler. When such jobs strive for additional resources due to various parameters including disaster at grid site, the Meta scheduler interacts with the proposed component requesting for additional resources. The component in turn will invoke the Infrastructure service of cloud in line with the policies incorporated, and the jobs will migrated to the cloud resources or if needed the virtual machines allocated will be involved into the virtual organization. Once the job execution gets completed, the virtual machines are given back to the cloud. In case of data grid when resources require additional storage space, it will be acquired from cloud.
On the other hand when the cloud is filled up with virtual machines or storage, the cloud controller will contact the component, which gets additional resource for cloud from grid and the image gets deployed in the grid as virtual instances providing a business model to grid. This entire process is abstract to the user.