accelerator

README

Nexledger Accelerator

Nexledger Accelerator is a software component designed to improve the performance of a blockchain network, e.g. Hyperledger Fabric, in terms of transaction throughput. Accelerator enables the blockchain network to deal with explosive transaction requests from applications.

Accelerator receives transactions from clients on behalf of blockchain nodes and provides transaction acceleration in terms of TPS (Transaction Per Second) by classifying, aggregating, and routing the transactions to blockchain network. The current version of Accelerator is compatible with Hyperledger Fabric v1.4.

Getting Started

Prerequisites

• Go (1.11.0 or greater)
• Docker (17.06.2-ce or greater)
• Docker-compose (1.14.0 or greater)

Building Accelerator

Accelerator supports go module for dependency management. To build the executable, please simply execute go build.

$ go build cmd/accelerator.go

Running ping example

The ping example shows how to configure and run Accelerator. The example is placed in In examples/ping.

To bootstrap Fabric network, please run start.sh script. It boots up the Hyperledger Fabric network including install/instantiation of the example chaincode.

$ ./examples/ping/start.sh

To serve requests from clients, Accelerator should be up and running with proper configuration.

$ ./accelerator -f examples/ping/configs/accelerator.yaml

Accelerator is a gRPC server, and the gRPC services are described in protos/accelerator.proto. You may send transactions using examples/ping/ping_test.go that has gRPC client for the ping example.

$  cd examples/ping
$  go test

You can terminate and remove the network by run stop.sh script.

(change dicrectory to root)
$ ./examples/ping/stop.sh

Under the hood

Modifying chaincode

Accelerator aggregates multiple transactions into a batched transaction and submits the batched transaction to the endorsers. Therefore, chaincodes operating with Accelerator should be modified to individually execute aggregated transactions.

contracts/src/ping/ping.go is the example chaincode with simple KV write/read operations. ping.go imports batchutil.go for segregating batched transactions from Accelerator and delegates the invocations to Invoke() in batchutil.go.

func (t *PingPongChaincode) Invoke(stub shim.ChaincodeStubInterface) pb.Response {
	fnc := string(stub.GetArgs()[0])
	switch fnc {
	case "ping":
		return Invoke(stub, t.ping)
	case "pong":
		return Invoke(stub, t.pong)
	}
	return shim.Error("Unknown action, check the first argument, must be one of 'insert', 'query'")
}

Configuring Accelerator

Accelerator should be configured with target chaincode functions and corresponding batch configs. The configuration file for ping example is placed at configs/accelerator.yaml

sdk: "examples/ping/configs/accelerator-sdk.yaml"
host: "localhost"
port: 8090
userName: "Admin"
organization: "peerorg1"
batch:
  - type: "execute"
    channelId: "accelerator"
    chaincodeName: "ping"
    fcn: "ping"
    queueSize: 1000
    maxWaitTimeSeconds: 5
    maxBatchItems: 10
  - type: "query"
    channelId: "accelerator"
    chaincodeName: "ping"
    fcn: "pong"
    queueSize: 1000
    maxWaitTimeSeconds: 5
    maxBatchItems: 10

• sdk: Path to the fabric SDK configuration File
• host: Host address of Accelerator
• port: Port number of Accelerator
• queueSize: The size of the in-memory queue that kept requested transactions until processing.
• maxWaitTimeSeconds: Maximum waiting time in seconds to create a new batch transaction.
• maxBatchItems: Maximum number of items for a new batch transaction.

Whitepaper

Whitepaper includes:

• The key design features of Accelerator enabling high performance enterprise-wide blockchain technology
• The evaluation results that show the performance improvement of Hyperledger Fabric by Accelerator in practical scenarios
• The use cases that provide an insight for understanding industrial blockchain platforms

Further Information

For further information please contact Samsung SDS(nexledger@samsung.com).