In the past decade, we have witnessed the rise of blockchain systems, encouraged by the hype around cryptocurrencies. Blockchain-based systems are capable of storing and sharing information while protecting its integrity--a property that distributed database systems provided for decades in different forms.

The authors of this paper approach the similarities between blockchains and database systems from the perspective of enabling mechanisms of both types of systems. Existing blockchain systems utilize many mechanisms borrowed from database know-how already. Still, the authors identify some key but not-so-obvious mechanisms to improve blockchain systems further. They take one of the most popular open-source blockchain systems, Fabric, and evaluate its transaction pipeline. Initial evaluations reveal that the system has a 100 percent success rate for blank transactions while it aborts the majority of meaningful transactions. By applying transaction reordering and early transaction abort, the authors are able to improve the throughput of Fabric significantly.

The paper systematically:

(1) Lays out the architecture and workflow used by Fabric;

(2) Identifies possible optimization mechanisms;

(3) Points out relevant similarities and shared technologies between Fabric and distributed database systems; and

(4) Discusses, in light of these findings, the optimization mechanisms applied to improve Fabric’s throughput.

Using Smallbank and a custom benchmark, the experiments show that the modifications on the Fabric system significantly increase the number of transactions per second and cut transaction latency by up to half of what it was. A peek at the individual effects of transaction reordering and early transaction abort reveals that, while each adaptation of the method improves the throughput individually, they make a significant impact when applied together.

Thus, the authors present interesting results with an impressive presentation. The paper makes for a compelling read while providing readers with substantial know-how on how blockchain transaction mechanisms work and how database research can be applied to these systems. The paper is especially interesting for the database community and researchers working on verified transaction systems.