TPS drops when scaling peers in existing organizations

[Snowball84]

I'm considering performance of hlf network(v 2.4). I first deployed a channel of 2 orgs, 1peer per org. Then I scaled to 2 peers per org, used hyperledger caliper to test TPS of both and found that TPS surprisingly drops (about 15%)when peer number increases. According to my understanding, increasing network size will load balancing endorsement between peers in an org and by no means decrease TPS. I'm using asset-transfer-basic' s update method for the test and MVCC_READ_CONFILCT error is avoided. I wonder whether this phenomenon makes sense and if not, how do I diagnose the performance bottleneck.

[pfi79]

I'm considering performance of hlf network(v 2.4). I first deployed a channel of 2 orgs, 1peer per org. Then I scaled to 2 peers per org, used hyperledger caliper to test TPS of both and found that TPS surprisingly drops (about 15%)when peer number increases. According to my understanding, increasing network size will load balancing endorsement between peers in an org and by no means decrease TPS. I'm using asset-transfer-basic' s update method for the test and MVCC_READ_CONFILCT error is avoided. I wonder whether this phenomenon makes sense and if not, how do I diagnose the performance bottleneck.

version 2.4 is not supported. please check 2.5 or 3.0 (with and without bft)