Tracking highly available peers in KAD

KAD is the largest DHT system with several million simultaneous users. Available peers play an important role in its healthy operations. In this paper, we performed an active measurement study to track highly available peers in KAD. Our measurement results show that the probability of KAD ID appearance in routing tables can be used to characterize peers´ availability. When a peer is online with a longer duration, the probability of its KAD ID appears in other peers´ routing tables increases, due to the message exchanges among peers. We characterize different types of KAD IDs in routing tables and find that the repeat times of KAD IDs follows the Pareto Law during one crawling session. In particular, we take some contrastive measurements to show the availability of the peers with high repeat times of KAD ID appearance in routing tables are much higher than the normal ones. At the end of this paper, we propose a simple way to improve lookup performance by using highly available peers in KAD.