D6.2 Phase 1 experiments


PRISTINE’s deliverable D6.2 is already available from the website’s deliverables section.

Executive summary

The first round of experiments carried out as part of the validation work of the PRISTINE SDK has shown promising results. Moreover, it also helped in identifying development issues that have been either corrected during this first development phase or that have been included in the development roadmap of RINA and the PRISTINE SDK for the second phase. Each experiment has covered different research areas (from resilient routing or aggregate congestion control to management aspects) and several RINA policies within each area.

Data-centre congestion control experiments: DIF Configuration

Data-centre congestion control experiments: DIF Configurationion

Several of the RINA benefits demonstrated during this first cycle of experiments apply to all three use cases. For example, the Loop Free Alternate (LFA) experiments can be always used by the network administrator to increase IPC service availability. Similarly, Explicit Congestion Notification (ECN) and vertical push-back between layers enable a faster and more efficient response to congestion than the prevalent end-to-end implicit feedback approach used in the Internet. ECN allows congestion effects to be confined closer to where congestion happens instead of affecting the whole network: each DIF manages the congestion in its own resources. This allows network providers to utilize network resources more efficiently and to run networks at higher utilization.

The development, debugging and testing of policies associated to the first cycle of experiments uncovered a number of bugs in the RINA implementation as well as some changes required in the SDK (which have already been addressed and released in the most up to date development branch, pristine-1.3). There are still open issues currently reported that will be dealt with during the second phase, mainly related to design improvements or non blocking bugs.

As far as business aspects are concerned, we have performed a preliminary qualitative analysis based on single experiments and how the KPIs impact on the business process of the use cases.

The work towards the final evaluation of the PRISTINE project results will focus on two main axes. On one hand, the research areas will continue to design and carry out experiments that further validate the usefulness of the PRISTINE SDK as far as RINA benefits are concerned. On the other hand, we intend to combine several research directions, designing and executing more complex experiments that help validate the SDK and RINA at a larger scale. We will define experiments that gather representative aspects of the three use cases, making sure that the simultaneous use of policies belonging to different research areas do not interfere with one another, but rather show at least the same level of performance they did when they were tested individually. These experiments will provide further feedback towards development regarding the combination of complex scenarios for RINA and the SDK, such as the use cases require.

Finally, we intend to carry out a higher level study focusing on the technical impact per use case and translate the results into the expected business impact.