Affects Version/s: None
Fix Version/s: 1.6.78
Sprint:EV 18.17 Service Pack, EV 18.18 Service Pack 2, EV 18.19
We need to prove that it will be stable under conditions similar to production use.
Perform a test of an Indy network that has the following attributes:
- The ledger is pre-loaded with 1 million transactions
- Pool size is 25 nodes.
- 1K concurrent clients
- Over a 3 hour period induce a sustained throughput of 10 write transactions per second and 100 read transactions per second on average (from clients)
Write load is a mixture of:
- writing credentials schema (5%),
- writing credential definition (5%)
- revoke registry definition (5%)
- revoke registry update (5%)
- write DID to ledger (20%)
- write payment to ledger (45%)
- write attrib to ledger (15%)
Read load is a mixture of:
- read DID from ledger (45%)
- read credential schema (10%)
- read credential definition (10%)
- read revoke registry definition (10%)
- read revoke registry delta (10%)
- read attrib from ledger (10%)
- read payment balance from ledger (5%)
- Write response time should be less that 5 seconds (would also like a report of the average).
- Read response time should be less than 1 second (would also like a report of the average).
- Unless view change in progress, pool should write 10 txns/sec and read 100 txns/sec
Main cases for research:
1) mixed set (imitation of active production work) - 1000 clients / 10 writes / 100 reads
2) payments only (understanding of pluggable ledgers productivity) - 1000 clients / as much as possible writes (1 - 4 - 10) / as much as possible reads (up to 10 - 40 - 100)
3) not payment transactions (comparision with historical productivity data) - 1000 clients / 10 writes / 100 reads