On Demand Resources
Imagine sitting at a restaurant, ordering food, and getting the drink in 5 minutes and have to wait for two days to get the meal served to you. In the case of on demand resources in IT and Telecom, you may receive some resources immediately and have to wait for others for a while, that is when you simply loose your appetite
On Demand IT and Network
Looking at the most dominant technologies in virtualisation, we see virtualisation in small and big datacenters as a foundation for on demand infrastructure, software and platforms. When it comes to network, we also see virtualisation in networking is dominant, creating private networks, VLANS and VPN facilitating international businesses. Now consider you want both IT and Network resources on demand, there is still work to be done to make this solution available. This could be due to lack of business incentives or lack of unified management system for IT and network environment. In 2013 we finished an EU projects - a brilliant group of researchers which focused on this topic.
Still to this date, datacenters are localised or there is only a network between the datacenters from the same provider. In most cases in far distance from the end user. This of course would impact the quality of experience for the end user. If you have your storage and computing resources in a data center, you have to face the delay and congestion in the IP backbone to reach your day to day data, and that is not really acceptable specially in case of time critical applications.
Cloud Networking should bring not only network as a virtualised resource, but also the possibility of distributed cloud, which enhances the quality of experience, dynamic connectivity, flexible consumption of resources, minimisation of IT investment, and reduction in IT costs, providing on demand networking in the same timeframe as IT resources.
We had two obvious use cases in mind. The first usecase was dynamic enterprise which introduced dynamic flexible network provisioning into the equation of on-demand and pay per use of IT solutions. The second usecase was the elastic video in the cloud, which could offer video and similar services from a cloud network ecosystem, leveraging on distributed computational resources at the edge of the network architecture, where cloud resources were geographically scattered inside the operator network.
In the dynamic enterprise scenario, we considered multiple cloud sites, implementing virtual processing and storage infrastructure, connected by an operator network. Some of these sites could belong to the enterprise and some could be from external public cloud providers, the operator network implemented the connectivity between the infrastructures implemented by the cloud sites and end users and could also provide processing and storage elements of its own. They collectively should implement a distributed virtual infrastructure solution, virtual infrastructure could be deployed on demand as a cloud network.
In the case of distributed cloud, elastic video delivery, cloud resources were geographically scattered inside the operator network in a more fine-grained fashion than traditional centralised data centres. The motivation was to leverage on the closeness of processing to the end users to offer enhanced quality of experience. In principal, this closeness implies that latency can be reduced and network bandwidth usage patterns can be optimised, by trading off between connectivity, processing and storage.
In such usecases obviously there are multiple administrative domains, it could be data center providers or network and connectivity providers ( network operators). When a high level resource requirement came in, the requirement could be delegated through delegation service control model, propagating it across service providers, which could finally get accommodated by one local resource management.
An important highlight was the distributed coordination across multiple administrative domains, which implied actual connections between different administrative domains, at lowest level, and even coordination at layer 2 and layer 3 of the network.
Starting the project, we had to take into the account, multiple provider, limited information disclosure, rapid provisioning, scale with number of providers as well as the number of users, while providers needed to be able to keep their autonomy and make their own decisions. In the architecture we also needed to define roles, interfaces and the concept of delegation and coordination across multiple operator domains.
In the case of dynamic enterprise, imagine if an enterprise has its main infrastructure deployed in one data center in Sweden and wants to attract customers in France. It can start to ask for infrastructure available in another data center and at the same time ask for the on demand connectivity to have all its infrastructure on the same network.
The project built the concepts and prototyped the components for on-demand bundling of computing, storage and networking resource via the same interface and in the same time frame.