Today's Internet provides no a priori guarantees about when or if packets will be delivered; this best-effort design has been key to Internet simplicity and scalability. Unfortunately, the Internet also provides no feedback on the fate of transmitted packets. As a result, when a network failure causes packets to get lost or delayed, the sender of the packets has no way of knowing where the failure occurred. This makes it hard for the sender to adapt to the failure (e.g. send its traffic via an alternative path) or hold any administrative entity accountable for it. End-hosts try to address this problem by using probing tools like trace routes. These tools, however, are unreliable, cannot be used to hold ISPs accountable, and are frequently blocked by ISPs for policy or security reasons.
This project proposes a solution, in which ISPs are motivated to willingly provide feedback on the loss and delay incurred by traffic within their network. Feedback is collected by monitoring boxes located on inter-ISP links and provided to interested parties. Our proposal attempts to reconcile the academic and industry perspectives on the Internet by developing a solution that is not only technically feasible, but also consistent with the current ISP business model: we enable ISPs to choose the level of feedback they provide, without revealing their internal structure.
The proposed mechanism reveals to senders which ISPs are accountable for losing or delaying their packets; as a result, senders are able to evaluate the performance of their ISPs and route their traffic around failure points. Unlike traditional quality-of-service solutions, we do not provide senders with a priori guarantees; we only provide them with a posteriori feedback, which senders can then use to adapt to failures. In essence, we try to improve the end-to-end performance of Internet communications in the face of failures, without compromising the best-effort nature of the Internet architecture.
Call for proposal
See other projects for this call