note that performance degrades as the blockheight increases. this wasn't sustainable after a few performance runs
also fixed-tps still performs better than transaction-load
As only 1 channel could not get peer above ~65% cpu, locks within the peer appear to restrict the cpu from utilising the full capacity. More channels being used will allow the cpu to be used more
Results of a theoretical maximum
but I managed better than this, but can't find the output
Results of a blind write of ~100 byte asset, 2 org 1 peer each, 1 orderer, EP(Org1 or Org2) with block cutting parameters of block_cut_time: 1s block_size: 50 preferred_max_bytes: 512 KB
read/write asset (fixed TPS 2750 read 1, write 1, same asset) - 1,000,000 assets preloaded. failures are MVCC_READ_CONFLICTS which are expected
note that performance degrades as the blockheight increases. this wasn't sustainable after a few performance runs also fixed-tps still performs better than transaction-load As only 1 channel could not get peer above ~65% cpu, locks within the peer appear to restrict the cpu from utilising the full capacity. More channels being used will allow the cpu to be used more