Summary of current test-run

[ineiti]

I'm using the benchmark BenchmarkIntegration_CustomVotesScenario in performance_test.go to test how much we need to speed things up. Here some preliminary results and thoughts:

• According to state_of_smart_contract.md, there is only a single instance of any smart contract. So:
  • it will get slower the more elections are stored
  • the proxy will probably have to handle this and return only appropriate data structures
• on my macbook I can get about 200 ballots stored, shuffled, and decrypted. More than that fails in the shuffling
• just casting ballots shows a big slowdown: the 100th ballot takes 0.7s, the 200th 1.2s, the 300th 1.7s, the 400th 2.3s, the 999th 5.4s, so an increase in casting a new ballot of 0.5s per 100 existing ballots
• The 100th ballot with 3 nodes takes 650ms, with 7 nodes it takes 720ms, so there is no big increase when the number of nodes increases

Probable issues to open (waiting on some more investigations):

• speed up casting a ballot by a factor of 100: it should be below 1s for 5 elections and 10'000 votes each. -> Reducing time to cast a ballot #53
• make shuffling work with 1'000 and 10'000 ballots
• make sure the smart contract is smart enough that the proxy can correctly extract the different elections
• verify the decryption is actually stored, too, and possibly even the results. Else it will take clients a lot of time to calculate the results

Once this works, the next steps are:

• test all of this from the client side
• using a real network with network delays

[PascalinDe]

closing this issue to clean up issues for transfer