Cloud Resource Orchestration
The cloud computing paradigm is steadily shifting the course of computing away from physical hardware with locally managed software platforms toward virtualized, cloud-based services. This new paradigm, brought about by cloud computing, provides users with instant access to virtually unlimited computing resources and, at the same time, allows providers to deploy complex IT infrastructure as a service (IAAS).
There is a huge benefit from economies of scale and tremendous gains in multiplexing offered by the sharing of the underlying physical infrastructure using virtualization. But all these benefits come at great cost. The sheer scale and highly dynamic nature of available cloud platforms bring considerable new challenges to service providers. To realize these complicated cloud services, a control framework must be present which can orchestrate cloud resource configuration, provisioning, utilization and decommissioning, and real-time monitoring across extensively distributed physical resources.
The main goal of cloud resource orchestration is the successful hosting and delivery of cloud-based applications and services through achieving the quality of service (QoS) objectives of users and clients. QoS is composed of functional and non-functional attributes such as security, consistency, reliability, integrity, availability, statistics, scalability and cost.
II. Challenges in Resource Orchestration
Cloud resource orchestration is becoming challenging because of the increasing sophistication of services provided through the cloud, the rapid growth of datacenters and the high failure rate of commercial hardware. Dynamic service features require the orchestration of resources across the different domains of computation, storage and networking, across geographically separated datacenters and in a sequential manner. Each of these operations is fairly complex and highly susceptible to failure. The final hurdle of a resource orchestration platform is concurrently handling numerous similar orchestration requests.
Resource orchestration functions must be performed while dealing with huge numbers of simultaneous user requests, enforcing engineering rules and service policies, and performing error handling, all in a very dynamic environment.
III. Approaches to Cloud Resource Orchestration
A. The Data-centric (DMF) Approach
One approach to resource orchestration is the data-centric management framework (DMF). This approach aims to maintain a conceptually centralized repository for data of all managed resources, including computational, storage and network devices. Every resource object has two copies which represent its state: the primary copy in the physical layer (physical device) and the secondary copy in the logical layer (resides in memory).
B.The Transactional Resource Orchestration Platform In the Cloud (TROPIC) Approach
This is a transactional orchestration resource platform for Infrastructure-as-a-Service (IaaS) cloud infrastructures. This approach to orchestration of physical resources is transactional, which can guarantee atomicity, consistency, isolation, and durability (ACID) of cloud service operations. Procedures are performed as transactions using ACID properties. Transactional semantics ensure that errors have no effect and that concurrent orchestrations will not violate rules due to clean abstraction.
Implementing a cloud resource orchestration framework is not an easy task, and there are different ways to implement such a framework, depending on the nature of the cloud service being provided. But no matter which kind of orchestration is implemented, it must meet user expectations and prevent or at least handle errors fairly well, while providing simultaneous orchestration services. There are new approaches being developed as new problems arise, so the field is still developing and maturing. We have yet to fully realize all the benefits which cloud computing can offer.
By Abdul Salam