Tamas,

In my experience maximizing throughput requires running multiple clients. You should be able to at least double if not triple your throughput by running 16 clients - assuming you have the computational capacity on your laptop.
You may also want to try the trick mentioned here to see if it gives you a small boost: https://github.com/hyperledger/fabric/wiki/Go-Performance-Portal#morestack

Bishop Brock : Research Staff Member : IBM Master Inventor
IBM Research : +1 (512) 286-5453 [Office] : +1 (512) 568-9467 [Mobile]
http://researcher.ibm.com/researcher/view.php?person=us-bcbrock

Message: 1
Date: Thu, 2 Jun 2016 18:10:24 +0200
From: Tamas Blummer <tamas@...>
To: hyperledger-fabric@...
Subject: [Hyperledger-fabric] Performance measurements
Message-ID: <EAC6B9D8-293D-4210-91DF-56E24795B0B1@...>
Content-Type: text/plain; charset="utf-8"

We created a noop system chain code that does nothing but store an invoke transaction.
Since it is a system chain code written in go and deployed at startup, it does not require the grpc-docker roundtrip at invoke.

Nevertheless we were only able to push 400 tx/s (few hundred bytes of random tx data each) with no consensus module activate on a single node on a laptop.

Do you have a more promising result or an idea how to reach magnitudes higher rates?

TAMAS BLUMMER
?CHIEF LEDGER ARCHITECT?



+?36 1 883 0300
?tamas@... <http://digitalasset.com/>
digitalasset.com <http://digitalasset.com/>