60 likes | 127 Views
This paper by Song Yongchao addresses the maintenance process in Distributed Hash Table (DHT) systems focusing on responsible peer replication. It covers steps for node failure, responsible peer failure, and graceful transitions, ensuring seamless data replication and management in DHT networks.
E N D
Responsible Peer(RP) Replicationin DHT level Song Yongchao (melodysong@huawei.com)
Step1: Choose location to replicate Choose peer A, B to replicate a keyM/value pair RP(responsible peer) maintenance maintenance B A
Step2: when replica nodes fail When one replica node fails, RP choose a new node to replicate RP Detect failure X X maintenance maintenance X B A maintenance New replica node C
Step3: when responsible peer fails RP X New RP for KeyM Detect failure X maintenance DHT lookup and transfer B A Replica peer lookup the new RP with looking up KeyM in the DHT and transfer KeyM/value to the new RP for KeyM
Step4: when the RP changes gracefully Handover keyM/value pair and replica Management to the New RP RP New RP for KeyM maintenance maintenance B A Due to overlay topology changes, the reponsible peer may change, RP transfers keyM/value to its new responsible peer and hand ove the management of replication for keyM to it
Step5: handle requests when responsible peer is busy RP Busy Relay the request to replica node for KeyM Get request for KeyM B A When the responsible peer is busy, it can relay requests of its responsible key to the replica, which can prevent its busy status getting worse.