Optimal Endorsement Policy for network-wide distributed blockchains

Blockchain is a technology that allows storing data in a secure and decentralized way. One of the main reasons why the interest in this technology is growing is because it can provide trust between untrusted parties without the use of a centralized entity. Hyperledger Fabric is a permissioned blockchain project maintained by The Linux Foundation meant for environments in which system performance is a key factor. For instance, HF architecture faces three consecutive phases called Execute, Order and Validate. This behavior differs from other blockchain architecture like Bitcoin in which an Order-Execute-Validate procedure is followed. During the Execution phase, Simulation happens: a proposal of the transaction, called Endorsement request, is sent to specific peers named Endorser peers or EPs. EPs execute the received endorsement request and send the obtained output back to the client in a message called Endorsement. This phase is called Simulation because, while executing the transactions, EPs don’t change the content of the data they are securely storing. The client needs to receive the Endorsement results from a combination of the above-mentioned EPs defined by a policy called Endorsement Policy. This policy is defined in a way that reflects the agreements taken between the participating organizations. This thesis describes Blockchain technology, HF architecture, and Fabric SDK, which is a Development Kit that enables the development of applications that interact with a Fabric network. Afterward, it is presented an optimal procedure of spreading the endorsement requests across the EPs to reduce the overall latency of the Endorsement procedure. This is done through an algorithm called OPEN that keeps track of the endorsement delay history of the network’s EPs. While respecting the Endorsement Policy, OPEN chooses the peers that performed better in the past transactions because they are the most likely to be the fastest in the next one. To keep updated the response delay history of not selected EPs, OPEN sends endorsement requests to additional peers. This will facilitate the better making of assumptions about the peers to endorse during the next transactions. Finally, having verified that the Endorsing Peers can queue requests when higher loads of transactions are submitted, an evaluation of OPEN performances in choosing the least congested EPs is carried out in compliance with other endorsing peer selection algorithms.