Empty Contract
The Empty Contract Benchmark consists of submitting emptyContract gateway transactions for the fixed-asset smart contract deployed within LevelDB and CouchDB networks. When submitting emptyContract gateway transactions, the interaction is recorded on the ledger. This results in the transaction pathway as depicted in Figure 1.
Figure 1: Submit Empty Contract Transaction Pathway
This is repeated for networks that use the following endorsement policies:
- 1-of-any
- 2-of-any
Achievable throughput and associated latencies are investigated through maintaining a constant transaction backlog of 15 transactions for each of the test clients.
Resource utilization is investigated for fixed TPS rate of 350TPS.
Benchmark Results
LevelDB- submit transactions with varying endorsement policy
| Type | Policy | Max Latency (s) | Avg Latency (s) | Throughput (TPS) |
|---|---|---|---|---|
| submit | 1-of-any | 0.41 | 0.09 | 485.4 |
| submit | 2-of-any | 0.33 | 0.10 | 420.0 |
CouchDB- submit transactions with varying endorsement policy
| Type | Policy | Max Latency (s) | Avg Latency (s) | Throughput (TPS) |
|---|---|---|---|---|
| submit | 1-of-any | 0.52 | 0.11 | 380.5 |
| submit | 2-of-any | 0.32 | 0.13 | 3387 |
LevelDB Resource Utilization– Submit By Policy @750TPS
CouchDB Resource Utilization– Submit By Policy @750TPS
Resource Utilization– Submit 1ofAny Policy @750TPS
Resource Utilization– Submit 2ofAny Policy @750TPS
Benchmark Observations
LevelDB is observed to be beneficial for achievable throughput and reduced latencies in comparison to CouchDB during submission of an emptyContract gateway transaction for both investigated endorsement policies.
With a fixed world state database, the endorsement policy is observed to impact the consumed resources when submitting a transaction. Increasing the number of required endorsements is observed to increase the CPU and network I/O, through inclusion of additional peers and smart contract containers required to participate in each transaction.
In comparing a LevelDB world state database with a CouchDB equivalent, only the network I/O is observed to be equivalent when varying the endorsement policy. There is an observed penalty in additional memory, CPU and disc I/O requirements for the use of a CouchDB world state for the network as a whole, though the memory requirements of the peers are reduced.