Skip to content

Latest commit

 

History

History
201 lines (145 loc) · 8.78 KB

README.md

File metadata and controls

201 lines (145 loc) · 8.78 KB

Morpheo Orchestrator Chaincode

This is the Orchestrator of the Morpheo platform with the blockchain. We use the private and permissioned solution called Hyperledger Fabric. Morpheo chaincode corresponds to the set of smart contracts, which are used to orchestrate operations on the Morpheo platform. It is the translation of Morpheo Orchestrator with a blockchain solution.

Licence: CECILL 2.1 (compatible with GNU GPL)

How to interact with the orchestrator

Use the Morpheo-Fabric-Bootstrap to create a network to interact with the Orchestrator. Once the network is up, the chaincode is installed and instantiated, you can go inside the docker cli to interact with the Orchestrator. Below some interaction examples, do not forget to set the correct environment variable:

peer chaincode query -n mycc -c '{"Args":["queryObject", "algo_1"]}' -C $CHANNEL_NAME
peer chaincode query -n mycc -c '{"Args":["queryObjects", "algo"]}' -C $CHANNEL_NAME
peer chaincode invoke -o orderer.morpheo.co:7050 --tls true --cafile $ORDERER_CA -n mycc -c '{"Args":["registerProblem", "e568587d-572c-4714-8084-378ed50d1c52", "2", "0pa81bfc-b5f4-5ba2-b81a-b464248f02a1, 0kk81bfc-b5f4-5ba2-b81a-b464248f02e3"]}' -C $CHANNEL_NAME
peer chaincode invoke -o orderer.morpheo.co:7050 --tls true --cafile $ORDERER_CA -n mycc -c '{"Args":["registerItem", "algo", "0pa81baa-b5f4-5ba2-b81a-b464248f02d2", "problem_1", "mytopalgo"]}' -C $CHANNEL_NAME
peer chaincode invoke -o orderer.morpheo.co:7050 --tls true --cafile $ORDERER_CA -n mycc -c '{"Args":["registerItem", "data", "9pa81bfc-b5f8-5ba2-b81a-b464248f02d2", "problem_1", "psg"]}' -C $CHANNEL_NAME
peer chaincode query -n mycc -c '{"Args":["queryProblemItems", "data", "problem_1"]}' -C $CHANNEL_NAME
peer chaincode query -n mycc -c '{"Args":["queryStatusLearnuplet", "todo"]}' -C $CHANNEL_NAME
// replace algo_0 with correct key
peer chaincode query -n mycc -c '{"Args":["queryAlgoLearnuplet", "algo_0"]}' -C $CHANNEL_NAME
// replace learnuplet_0 with correct key
peer chaincode invoke -o orderer.morpheo.co:7050 --tls true --cafile $ORDERER_CA -n mycc -c '{"Args":["setUpletWorker", "learnuplet_0", "Arbeiter_12"]}' -C $CHANNEL_NAME
// replace learnuplet_0 with correct key
peer chaincode invoke -o orderer.morpheo.co:7050 --tls true --cafile $ORDERER_CA -n mycc -c '{"Args":["reportLearn", "learnuplet_0", "done", "0.82", "{\"data_3\": 0.78, \"data_4\": 0.88}", "{\"data_2\": 0.80}"]}' -C $CHANNEL_NAME

Chaincode-docker-devmode

You can use the chaincode-docker-devmode to more easily develop the chaincode, as detailed here

Smart contracts documentation

Elements of the ledger

The ledger is a key value store. To be able to make complex queries, such as querying all algorithms related to a problem, we use CompositeKey.

We call an ObjectType a type of element of the ledger (similar to a table in a relational database).

Data and Algo

Data and algo are 2 ObjectTypes, which both derive from an Item structure:

type Item struct {
    ObjectType     string `json:"docType"`
    StorageAddress string `json:"storageAddress"`
    Problem        string `json:"problem"`
    Name           string `json:"name"`
}

Keys: data_<uuid> and algo_<uuid>. Associated composite keys: data~problem~key and algo~problem~key.

Problem

A problem derives from the Problem structure:

type Problem struct {
    ObjectType       string   `json:"docType"`
    StorageAddress   string   `json:"storageAddress"`
    SizeTrainDataset int      `json:"sizeTrainDataset"`
    TestData         []string `json:"testData"`
}

Keys: problem_<uuid>.

Learnuplet

A learnuplet derives from the Learnuplet structure:

type Learnuplet struct {
    ObjectType        string             `json:"docType"`
    Problem           map[string]string  `json:"problem"`      // {problemKey: problemStorageAddress}
    Algo              map[string]string  `json:"algo"`         // {algoKey: algoStorageAddress}
    ModelStartAddress string             `json:"modelStartAddress"`
    ModelEndAddress   string             `json:"modelEndAddress"`
    TrainData         map[string]string  `json:"trainData"`    // {data1Key: data1StorageAddress, ...}
    TestData          map[string]string  `json:"testData"`     // {data1Key: data1StorageAddress, ...}
    Worker            string             `json:"worker"`
    Status            string             `json:"status"`
    Rank              int                `json:"rank"`
    Perf              float64            `json:"perf"`
    TrainPerf         map[string]float64 `json:"trainPerf"`
    TestPerf          map[string]float64 `json:"testPerf"`
}

Keys: learnuplet_<uuid>. Associated composite key: learnuplet~algo~key.

Smart Contracts

+ queryObject: to query a given object

Args:

  • objectKey, such as data_8fa81bfc-b5f4-4ba2-b81a-b464248f02d3, learnuplet_ca3a5a53-9684-429f-9896-4f7c94f9def0
peer chaincode query -n mycc -c '{"Args":["queryObject", "learnuplet_ca3a5a53-9684-429f-9896-4f7c94f9def0"]}' -C $CHANNEL_NAME

+ queryObjects: to query objects of a given type

Args:

  • objectType, such as data, learnuplet
peer chaincode query -n mycc -c '{"Args":["queryObjects", "learnuplet"]}' -C $CHANNEL_NAME

+ queryProblemItems: to query data or algos related to a problem

Args:

  • itemType: data or algo
  • problemKey, such as problem_8fa81bfc-b5f4-4ba2-b81a-b464248f02d3
peer chaincode query -n mycc -c '{"Args":["queryProblemItems", "data", "problem_8fa81bfc-b5f4-4ba2-b81a-b464248f02d3"]}' -C $CHANNEL_NAME

+ registerItem: to register an algo or data

TODO: modify function to register several data at a time

Args:

  • itemType: data or algo
  • storageAddress, for now it corresponds to the uuid on Storage, such as 0pa81bfc-b5f4-5ba2-b81a-b464248f02d2
  • problemKey, such as problem_2
  • itemName, such as mysuperalgo
peer chaincode invoke -o orderer.morpheo.co:7050 --tls true --cafile $ORDERER_CA -n mycc -c '{"Args":["registerItem",
"algo", "0pa81bfc-b5f4-5ba2-b81a-b464248f02d2", "problem_1", "topalgo"]}' -C $CHANNEL_NAME
peer chaincode invoke -o orderer.morpheo.co:7050 --tls true --cafile $ORDERER_CA -n mycc -c '{"Args":["registerItem", "data", "9pa81bfc-b5f4-5ba2-b81a-b464248f02d2", "problem_1", "psg"]}' -C $CHANNEL_NAME

+ registerProblem: to register a new problem

Args:

  • storageAddress: address of the problem workflow on storage
  • sizeTrainDataset: number of train data per mini-batch
  • testData: list of test data adresses on storage
peer chaincode invoke -o orderer.morpheo.co:7050 --tls true --cafile $ORDERER_CA -n mycc -c '{"Args":["registerProblem", "dda81bfc-b5f4-5ba2-b81a-b464248f02d2", "2", "0pa81bfc-b5f4-5ba2-b81a-b464248f02a1, 0pa81bfc-b5f4-5ba2-b81a-b464248f02e3"]}' -C $CHANNEL_NAME

+ queryStatusLearnuplet: to query all learnuplets with a given status

Args:

  • status: todo, pending, failed, or done
peer chaincode query -n mycc -c '{"Args":["queryStatusLearnuplet", "todo"]}' -C $CHANNEL_NAME

+ queryAlgoLearnuplet: to query all learnuplets associated to a given algo

Args:

  • algoKey: algo key of the algo of interest
peer chaincode query -n mycc -c '{"Args":["queryAlgoLearnuplet", "algo_f50844e0-90e7-4fb8-a2aa-3d7e49204584"]}' -C $CHANNEL_NAME

+ setUpletWorker: to set the worker and change the status of a learnuplet

Args:

  • learnupletKey, such as learnuplet_f50844e0-90e7-4fb8-a2aa-3d7e49204584
  • worker: worker identifier... to be defined
peer chaincode invoke -o orderer.morpheo.co:7050 --tls true --cafile $ORDERER_CA -n mycc -c '{"Args":["setUpletWorker", "learnuplet_f50844e0-90e7-4fb8-a2aa-3d7e49204584", "Arbeiter_12"]}' -C $CHANNEL_NAME

+ reportLearn: to report the output of a learning task

Args:

  • learnupletKey, such as learnuplet_f50844e0-90e7-4fb8-a2aa-3d7e49204584
  • status: done or failed
  • perf: performance of the model (performance on test data), such as 0.99
  • trainPerf: performances on each train data, such as {\"data_12\": 0.89, \"data_22\": 0.92, \"data_34\": 0.88, \"data_44\": 0.96}
  • testPerf: performances on each test data, such as {\"data_2\": 0.82, \"data_4\": 0.94, \"data_6\": 0.88}
peer chaincode invoke -o orderer.morpheo.co:7050 --tls true --cafile $ORDERER_CA -n mycc -c '{"Args":["reportLearn", "learnuplet_f50844e0-90e7-4fb8-a2aa-3d7e49204584", "done", "0.82", "{\"data_3\": 0.78, \"data_4\": 0.88}", "{\"data_2\": 0.80}"]}' -C $CHANNEL_NAME

TODO

  • register several data at a time