ACTIVE ANTI-ENTROPY MECHANISM BASED ON SPECTRAL BLOOM FILTER AND PH-2 HASH ALGORITHM FOR RECONCILATION OF REPLICAS OF NOSQL DISTRIBUTED DOCUMENT ORIENTED DATABASES

Abstract

Information systems are used in many areas of human activity, which are not limited to one country or continent. This may require horizontal scaling for the system to function properly. Ignoring this can affect performance and availability, which in turn can lead to a loss of reputation and users. Horizontal scaling increases the number of database replicas, which creates the need for data reconciliation, since writing operations to different nodes increases entropy. There are various technologies aimed at reducing it, including Active Anti-Entropy. Its essence is to detect inconsistencies and start the reconciliation process between replicas. It is actively used in a database such as Riak and uses the Merkle Tree data structure, which is based on the use of hashing algorithms. The speed of inconsistency identification depends on the chosen hashing algorithms and the number of documents in the collection. An increase in the number of documents or even their size can worsen the even distribution and lead to an increase in the number of collisions. The occurrence of collisions increases the time period of data inconsistency, because the system cannot detect the inconsistency in time. In addition to the collisions that can occur, you need to consider the delay due to data transfer over the network when nodes interact, and remember that such verification is not a one-time operation, but requires constant computation on replicas and sending for verification. Minimizing the time of these operations will speed up the data reconciliation process. Critically important data must be reconciled with minimal delay, as an untimely or incorrectly made decision can lead to material or even human losses. To prevent this, there must be a solution that will minimize the delay of matching such data.