IMPROVING DATA AVAILABILITY IN PEER-TO-PEER CONTENT-ADDRESSED SYSTEMS THROUGH PROACTIVE REPLICATION

Abstract

Modern decentralized systems, including IPFS, predominantly employ a “pull-on-demand” approach, in which replica- tion occurs only as a side effect of data retrieval. This leads to low availability in small networks where the number of active nodes is insufficient to naturally form an adequate number of copies. The paper examines an approach to improving data availability in peer-to-peer content-addressable systems through proactive replication and deterministic block placement. A mechanism is developed and analyzed that ensures predictable data availability without substantially increasing coor- dination among nodes and while preserving compatibility with DHTs. The requirements for a data-placement protocol in peer- to-peer networks are formulated, a proactive replication model, aligned with block distribution in the keyspace is proposed; and the capability of the approach to reduce load on individual nodes and achieve higher data availability is evaluated. Solution that integrates DHT-based routing with a placement mechanism driven by block identifiers that determine their position in the keyspace is proposed. Two-tier pinning model is introduced: hard pins guarantee persistent storage of blocks, whereas soft pins are created automatically during replication and have a limited lifetime that is extended upon repeated access. A garbage collector interprets combinations of pins to determine which blocks may be safely removed, retaining only relevant and frequently used data. Experimental results demonstrate that the proposed approach decreases the load on individual nodes and increases the system’s resilience to failures, making it a promising option for small private peer-to-peer networks with stringent availability requirements.