Lyatiss Blog
Information and opinions about Lyatiss and the cloud ecosystem.
The Cloud and the Problem of Best Effort Networking
The Cloud, or simply on-demand computing, is not just a new technology about to revolutionize information technology. It is revolutionizing businesses. With Cloud solutions, businesses and users are freed from the costly and time-consuming tasks of acquisition, installation and maintenance of servers, storage systems, network equipment and applications. With the Cloud, only a few clicks are needed to reserve and access computing infrastructure (IaaS: Infrastructure as a Service), development platforms (PaaS: Platform as a Service) and applications (SaaS: Software as a Service) over the Web.
The public Cloud provides on-demand access to the resources of an external provider, while a private Cloud provides on-demand access to the company’s internal resources over a private network. With a public or private Cloud, shared resources can be dynamically allocated to different applications, services or projects. The main advantage of this approach is the elasticity in provisioning equipment and computing capacity. To meet a peak in demand, a service can “borrow” resources from the shared pool of resources. When activity slows down, unused resources are returned to the pool.
Best Effort Networking for the Cloud means best-effort performance and security
The Cloud enables the ability to multiplex resources and share them across multiple client tenants, thus significantly enhancing resource utilization. For a service provider, there is a great opportunity for new revenue generation in the Cloud enabled by the economies of scale of hosting equipment and enabling application, platform or infrastructure access to a large numbers of customers. Resource sharing and capacity multiplexing are thus attractive propositions for businesses and operators alike.
This is why the Cloud has met with unprecedented enthusiasm. It is enabling immediate and cost-effective fingertip access to computing, development environments as well as applications. It is profoundly transforming our digital world.
One of the key enabling technologies for the Cloud has been server virtualization that allows physical servers to be sliced into pools of virtual servers that can used by applications on-demand. Current Cloud provisioning typically entails allocation of virtual machines and storage resources from a pool of resources to applications, as they need resources.
The backplane of the Cloud is the Internet, the network of networks. The Cloud relies on networking technology that is poorly adapted to the new requirements. Public and private Cloud users today access computing resources using a communications protocol suite, the Internet’s TCP/IP stack, offering a service level defined as “best-effort”. Data exchange within the Cloud itself also uses the same TCP/IP communication paradigms. Distributed processes synchronize themselves and communicate throughout a best-effort network in terms of performance but also in terms of security!
The consequences of best effort networking for the Cloud can be variable user experience, lack of security, non-optimal network utilization, and a need to over-provision resources.
Will users and businesses moving to the Cloud be satisfied with the current situation and the lack of control of security and end-to-end service quality? And as traffic grows and as a new generation of network sensitive Cloud applications are introduced will this problem only get worse?
One thing is clear – at this stage of the Cloud evolution a clean slate approach to networking is not feasible. There is a rich legacy of existing network devices and technologies that companies have invested in. Cloud networking that is on-demand, cost-effective and facilitates end-to-end application security and service quality will have to adapt to the deployment reality which is a mix of existing network technologies and emerging technologies.