Permissioned Blockchain Performance
See also
Enterprise Ethereum
Quorum
Arati Baliga, I Subhod, Pandurang Kamat, Siddhartha Chatterjee (Persistent Systems)
3 peers for Raft, 4 peers for IBFT.
RAFT and IBFT are comparable in terms of throughput except that RAFT performs slightly better at higher input transaction rates above 1650 tx/sec
https://gyazo.com/765f11f0fb66c1e5cdbcd26e175e927c
For private contracts, the maximum achievable load on the system was 900 tx/sec beyond which the system failed to reach consensus.
Hyperledger BESU
All in all, our efforts have moved us from 300 TPS on Besu 1.4.0 to 350 for Besu 1.5 and 400 TPS when using GraalVM. In order to get repeatable measurements for our TPS improvements, we worked with Hyperledger Caliper, another project under the Hyperledger umbrella to benchmark our numbers across various Besu versions
go-ethereum (geth)
Sara Rouhani (University of Saskatchewan), et al.
ICSESS'17
https://gyazo.com/560c522aa0185f0646d868589d7d3f37
BPM'19
https://gyazo.com/de7b37d9c280163993a1788fe720b31d
The first one (‘account’) changes the balance of two addresses, while the other one (‘ballot’) only accesses its own state
https://gyazo.com/a7b3084d6805280da757bc0eddc404db
Suporn Pongnumkul, Chaiyaphum Siripanpornchana, and Suttipong Thajchayapong (National Electronics and Computer Technology Center, Thailand)
ICCCN'17
geth and Fabric
https://gyazo.com/160323ea6c9fe2d86bad2d833b7b2049
STELLA - a joint research project of the European Central Bank and the Bank of Japan
Hyperledger Fabric Version 0.6.1
Yue Hao (Beijing University of Posts and Telecommunications), etl al.
https://gyazo.com/e324b6a1562d059dcec55924deb83c83
Pu Yuan, Kan Zheng, Xiong Xiong, Kuan Zhang, Lei Lei
Optimized
Christian Gorenflo, Stephen Lee, Lukasz Golab, S. Keshav
1. Separating metadata from data: Redesign Fabric’s transaction ordering service to work with only the transaction IDs (not whole transactions)
2. Parallelism and caching: Rredesign Fabric’s transaction validation service by aggressively caching unmarshaled blocks at the comitters and by parallelizing as many validation steps as possible, including endorsement policy validation and syntactic verification.
3. Exploiting the memory hierarchy for fast data access on the critical path: Fabric’s key-value store that maintains
world state can be replaced with light-weight in-memory data structures (hash table).
4. Resource separation: the peer roles of committer and endorser vie for resources. We introduce an architecture that moves these roles to separate hardware
Fully plug-and-play and do not require any interface changes
Parth Thakkar, Senthil Nathan, Balaji Vishwanathan
Improved the overall throughput by 16× (i.e., from 140 tps to 2250 tps)
Corda
DevDay: Corda Enterprise: Journey to 1000 TPS per node, Rick Parker Slides Nihhaar Chandra R
Others
Libra
Jiashuo Zhang, Jianbo Gao, Zhenhao Wu, Wentian Yan, Qize Wu, Qingshan Li, Zhong Chen (Peking University)
Result with P2P Transfer scirpt (1000 txs generated)
https://gyazo.com/b0d3f2e4ed1992b5836be4c02cbc1e46