A blockchain benchmark framework to measure performance of multiple blockchain solutions
Switch branches/tags
Nothing to show
Clone or download
aklenik Merge pull request #194 from nklincoln/guides
add issue and PR templates
Latest commit b9d4db2 Oct 19, 2018

README.md

Caliper Introduction

Caliper is a blockchain performance benchmark framework, which allows users to test different blockchain solutions with predefined use cases, and get a set of performance test results.

Currently supported blockchain solutions:

Hyperledger Composer is also supported, please see Composer Performance Test.

Currently supported performance indicators:

  • Success rate
  • Transaction/Read throughput
  • Transaction/Read latency(minimum, maximum, average, percentile)
  • Resource consumption (CPU, Memory, Network IO,...)

See [to add the link to PSWG] to find out the definitions and corresponding measurement methods.

Achitecture

See Architecture introduction.

Build

Pre-requisites

Make sure following tools are installed

  • NodeJS 8.X
  • node-gyp
  • Docker
  • Docker-compose

Run npm install in caliper folder to install dependencies locally

Install blockchain SDKs

  • Fabric

    • Install using the repository (for the supported Fabric v1.1)
      • run npm install grpc@1.10.1 fabric-ca-client@1.1.0 fabric-client@1.1.0 in the root folder
      • If you want to test Fabric with old version such as v1.0.0, you should install compatible client SDK,
        e.g. npm install grpc@1.10.1 fabric-ca-client@1.0.0 fabric-client@1.0.0
  • Sawtooth

    • Install dependencies

      $npm install protocol-buffers
      
    • Install sawtooth javascript sdk using repository

      • run npm install sawtooth-sdk in the root folder
  • Iroha

    • Install Iroha Library by npm install --no-save iroha-lib@0.1.7 in Caliper's root folder.
    • The package is in alfa phase, so if you have some problems with installing or compilation - please contact Iroha maintainers.
  • Composer

    • Install dependencies

    The easiest way to get started using a target version of Composer is to update the main package.json file to include the required Composer and Fabric modules, and subsequently run an npm install command. It is important that the Composer and Fabric versions are compatible.

     "composer-admin": "0.19.0",
     "composer-client": "0.19.0",
     "composer-common": "0.19.0",
     "fabric-ca-client": "1.1.0",
     "fabric-client": "1.1.0",
    

    Please see the plugin documentation for more details on using the Composer performance plugin, and developing your own tests.

Run benchmark

All predefined benchmarks can be found in benchmark folder. To start your first benchmark, just run this in root folder

node benchmark/simple/main.js -c yourconfig.json -n yournetwork.json
  • -c : specify the config file of the benchmark, if not used, config.json will be used as default.
  • -n : specify the config file of the blockchain network under test. If not used, the file address must be specified in the benchmak config file.

Some example SUTs are provided in network folder, they can be launched automatically before the test by setting the bootstrap commands in the configuration file, e.g

{
  "command" : {
    "start": "docker-compose -f network/fabric/simplenetwork/docker-compose.yaml up -d",
    "end" : "docker-compose -f network/fabric/simplenetwork/docker-compose.yaml down;docker rm $(docker ps -aq)"
  }
}

The scripts defined in command.start will be called before the test, and the scripts defined in command.end will be called after the finish of all tests. You can use them to define any preparation or clean-up works.

You can also run the test with your own blockchain network, a network configuration should be provided and corresponding file path should be specified in configuration file's blockchain.config.

Note:

  • When running the benchmark, one or more blockchain clients will be used to generate and submit transactions to the SUT. The number of launched clients as well as testing workload can be defined using the configuration file.
  • A HTML report will be generated automatically after the testing.

Alternative

You can also use npm scripts to run a benchmark.

  • npm run list: list all available benchmarks
$ npm run list

> caliper@0.1.0 list /home/hurf/caliper
> node ./scripts/list.js

Available benchmarks:
drm
simple
  • npm test: run a benchmark with specific config files
$ npm test -- simple -c ./benchmark/simple/config.json -n ./benchmark/simple/fabric.json

> caliper@0.1.0 test /home/hurf/caliper
> node ./scripts/test.js "simple" "-c" "./benchmark/simple/config.json" "-n" "./benchmark/simple/fabric.json"
......

Run benchmark with distributed clients (experimental)

In this way, multiple clients can be launched on distributed hosts to run the same benchmark.

  1. Start the ZooKeeper service

  2. Launch clients on target machines separately by running node ./src/comm/client/zoo-client.js zookeeper-server or npm run startclient -- zookeeper-server . Time synchronization between target machines should be executed before launching the clients.

    Example:

    $ npm run startclient -- 10.229.42.159:2181
    
    > caliper@0.1.0 startclient /home/hurf/caliper
    > node ./src/comm/client/zoo-client.js "10.229.42.159:2181"
    
    Connected to ZooKeeper
    Created client node:/caliper/clients/client_1514532063571_0000000006
    Created receiving queue at:/caliper/client_1514532063571_0000000006_in
    Created sending queue at:/caliper/client_1514532063571_0000000006_out
    Waiting for messages at:/caliper/client_1514532063571_0000000006_in......
  3. Modify the client type setting in configuration file to 'zookeeper'.

    Example:

    "clients": {
      "type": "zookeeper",
      "zoo" : {
        "server": "10.229.42.159:2181",
        "clientsPerHost": 5
      }
    }
    
  4. Launch the benchmark on any machine as usual.

Note:

  • Zookeeper is used to register clients and exchange messages. A launched client will add a new znode under /caliper/clients/. The benchmark checks the directory to learn how many clients are there, and assign tasks to each client according to the workload.
  • There is no automatic time synchronization between the clients. You should manually synchronize time between target machines, for example using 'ntpdate'.
  • The blockchain configuration file must exist on machines which run the client, and the relative path (relative to the caliper folder) of the file must be identical. All referenced files in the configuration must also exist.

Write your own benchmarks

Caliper provides a set of nodejs NBIs (North Bound Interfaces) for applications to interact with backend blockchain system. Check the src/comm/blockchain.js to learn about the NBIs. Multiple Adaptors are implemented to translate the NBIs to different blockchain protocols. So developers can write a benchmark once, and run it with different blockchain systems.

Generally speaking, to write a new caliper benchmark, you need to:

  • Write smart contracts for systems you want to test
  • Write a testing flow using caliper NBIs. Caliper provides a default benchmark engine, which is pluggable and configurable to integrate new tests easily. For more details, please refer to Benchmark Engine .
  • Write a configuration file to define the backend network and benchmark arguments.

Directory Structure

Directory Description
/benchmark Samples of the blockchain benchmarks
/docs Documents
/network Boot configuration files used to deploy some predefined blockchain network under test.
/src Souce code of the framework
/src/contract Smart contracts for different blockchain systems

How to contribute

See Contributing

License

The Caliper codebase is release under the Apache 2.0 license. Any documentation developed by the Caliper Project is licensed under the Creative Commons Attribution 4.0 International License. You may obtain a copy of the license, titled CC-BY-4.0, at http://creativecommons.org/licenses/by/4.0/.