Lead: ETH
This task will define the use cases that will be explicitly studied during the project to demonstrate the capabilities of mPlane. The definition of the use cases and requirements to be supported by mPlane will be the first technical action in the project. The use cases will be defined so as to exploit the technical features of the mPlane platform:
Already during the proposal preparation phase, the consortium has produced an initial list of candidate use cases to be addressed by the mPlane project. These are listed below and will be elaborated, refined and possibly extended within the course of the project.
Today, applications are moving off the desktop into the cloud, and this trend is expected to continue and accelerate. In the logical conclusion of this trend, End-User terminals become thin clients, little more than user interfaces to applications or even whole virtual machines full of applications running on the cloud, with data stored in the cloud. Similarly, CDN are today the vital part of the Internet. Most of the traffic is served by some CDN, and even cloud systems exploit CDN to alleviate load on data-centres. While this increases application deployment flexibility, maintainability, and data storage reliability, it also vastly increases the dependence of the application on the proper operation and performance of the network. Key to the end-to-end reliability of cloud computing is the ability of Application Providers and End-Users to diagnose performance and availability issues with cloud computing services. This use case will demonstrate the applicability of the mPlane approach to cloud computing reliability assurance.
Quality of experience is crucial for both emerging multimedia applications like Internet TV and videoconferencing, as well as for the evolution of more traditional web browsing into rich, interactive web content. Complex network interactions lead not only to hard-to-pinpoint causes for availability and performance issues in both application domains, but the link between traditional objectively measured quality of service (QoS) and user perceived QoE is not well understood. This use case will focus on root cause analysis of QoE issues which can be initiated by an Application Provider, an ISP, or an End-User; in the latter case, the focus will be on the development of a simple interface where a user can point out poor experience, and drill down to the cause or file a trouble ticket with their network provider for further assistance.
Service level agreements are an important line of business and revenue source for ISPs, wherein they guarantee given performance and downtime parameters to protect business-critical services of their customers. An application of mPlane could be certification and verification of these SLAs, which allows ISPs to prove they can meet a given SLA commitment during the sales process, enterprise customers and Regulatory Agencies to verify that SLA commitments are being met, and ISPs to troubleshoot the issue when they are not.
One of the key challenges for the research community is to define effective algorithms to automatically define the “normal” behaviour of the network/application/traffic/users and then devise algorithms to identify deviations from this normality, i.e., to identify “changes” in the patterns. In this context, the amount of information exposed by mPlane opens room to define novel approaches. In particular, the spatial and temporal diversity of data provided by mPlane can be exploited to define novel algorithms that automatically define the baseline behaviour and then trigger alarms in case of significant deviations.
Today’s cellular networks are increasingly used by a larger and larger number of smartphones, devices with the ability to generate and sink a previously unforeseen amount of network traffic. Smartphones offer a fertile environment for the development of a number of innovative applications, some of which tend to generate traffic even when not used by their owners (such as push service, or notifications from the cloud). Multitasking as well as the reliance of smartphones on intelligent backends make the debugging of smartphone applications very difficult. In this use case we will study the performance of a variety of web workloads under a variety of network scenarios with the aim to pinpoint performance degradation to the application, the device, the network, or the server.
Each of these use cases will consider the impact on each of the stakeholders in a given measurement; in the mPlane vision, each of these not only has an interest in the measurement to be performed (passive involvement), but also contribute to perform it (active involvement):