6.8 KiB
Overview
Rosetta-IO is a building block of Rosetta, providing IO service for Rosetta. Once the channel in Rosetta-IO is created, it is available to send/receive messages from other nodes. In Rosetta-IO, NODE identified by NODE ID, is a endpoint of a TCP connection. Each node has unique (HOST, PORT) tuple where HOST is a IP ADDRESS or DOMAIN NAME and PORT is a TCP LISTENING PORT.
Rosetta-IO supports MULTI DATA SOURCE. It defines three roles, called DATA ROLE, COMPUTATION ROLE and RESULT ROLE. DATA ROLE denotes that inputing data/module from the node is legal. Nodes, owning COMPUTATION ROLE, can participate in PRIVACY COMPUTATION. The results of PRIVACY COMPUTATION can only be stored on nodes with RESULT ROLE. Each node can have one or more roles. Rosetta-IO establishes network topoloy according to the roles of nodes. Three subnetworks, namely INPUT NETWORK, COMPUTATION NETWORK and OUTPUT NETWORK, will be built. INPUT NETWORK consists of nodes owning DATA ROLE and/or COMPUTATION ROLE. All of the nodes in COMPUTATION NETWORK have COMPUTATION ROLE. Nodes with COMPUTATION ROLE and nodes with RESULT_ROLE establish OUTPUT NETWORK together.
Rosetta-IO supports MULTI TASKING. The term CHANNEL refers to a task level handler. Every interface provided is based on a specific task. In Rosetta-IO, the connection between any two nodes is reused inter-task and there is only one connection between any two nodes.
Compile & Install
Follow the steps below to compile and install Rosetta-IO
$ git clone --recurse https://github.com/LatticeX-Foundation/Rosetta-IO.git
$ cd Rosetta-IO
$ export install_path=~/.local/rosetta-io
$ mkdir -p build && cd build
$ cmake ../ -DCMAKE_INSTALL_PREFIX=${install_path}
$ core_num=$(nproc)
$ make -j${core_num} && make install
Example
Here we give a example to demostrate how to use Rosetta-IO. Support three parties, called P0, P1 and P2, need to compare their config file on network topology. They may write the following program to finish this task.(check_config_json.cpp)
#include <stdio.h>
#include <string.h>
#include <string>
#include <vector>
#include <map>
#include <io/internal_channel.h>
#include <io/channel.h>
#include <io/channel_decode.h>
using namespace std;
int main(int argc, char* argv[]) {
if (argc != 3) {
printf("argc != 3\n");
return -1;
}
const char* file_name = argv[1];
const char* node_id = argv[2];
string config_str = "";
char buf[1024];
FILE* fp = fopen(file_name, "r");
if (fp == nullptr) {
printf("open file %s error", file_name);
return -1;
}
while (fgets(buf, sizeof(buf), fp) != nullptr) {
config_str += string(buf);
}
fclose(fp);
printf("config:%s", config_str.c_str());
IChannel* channel = ::CreateInternalChannel("test", node_id, config_str.c_str(), nullptr);
vector<string> connected_nodes = ::decode_vector(channel->GetConnectedNodeIDs());
int config_str_size = config_str.size();
const char* data_id = "config str";
for (int i = 0; i < connected_nodes.size(); i++) {
channel->Send(connected_nodes[i].c_str(), data_id, (const char*)&config_str_size, sizeof(int));
channel->Send(connected_nodes[i].c_str(), data_id, config_str.data(), config_str_size);
printf("send data to %s\n", connected_nodes[i].c_str());
}
for (int i = 0; i < connected_nodes.size(); i++) {
int data_size = 0;
channel->Recv(connected_nodes[i].c_str(), data_id, (char*)&data_size, sizeof(int));
string data(data_size, 0);
channel->Recv(connected_nodes[i].c_str(), data_id, &data[0], data_size);
printf("recv data from %s, size:%d\n", connected_nodes[i].c_str(), data_size);
}
::DestroyInternalChannel(channel);
return 0;
}
P0 and P1 use the following config file, named p0p1.json.
{
"NODE_INFO": [
{
"NAME": "PartyA(P0)",
"HOST": "127.0.0.1",
"PORT": 11121,
"NODE_ID": "P0"
},
{
"NAME": "PartyB(P1)",
"HOST": "127.0.0.1",
"PORT": 12144,
"NODE_ID": "P1"
},
{
"NAME": "PartyC(P2)",
"HOST": "127.0.0.1",
"PORT": 13169,
"NODE_ID": "P2"
}
],
"DATA_NODES": [
"P0",
"P1",
"P2"
],
"COMPUTATION_NODES": {
"P0": 0,
"P1": 1,
"P2": 2
},
"RESULT_NODES": [
"P0",
"P1",
"P2"
]
}
P2 uses the follow config file, named p2.config.
{
"NODE_INFO": [
{
"NAME": "PartyA(P0)",
"HOST": "127.0.0.1",
"PORT": 11121,
"NODE_ID": "P0"
},
{
"NAME": "PartyB(P1)",
"HOST": "127.0.0.1",
"PORT": 12144,
"NODE_ID": "P1"
},
{
"NAME": "(P2)",
"HOST": "127.0.0.1",
"PORT": 13169,
"NODE_ID": "P2"
}
],
"DATA_NODES": [
"P0",
"P1",
"P2"
],
"COMPUTATION_NODES": {
"P0": 0,
"P1": 1,
"P2": 2
},
"RESULT_NODES": [
"P0",
"P1",
"P2"
]
}
The only difference between the two config files is that NAME of P2 in p0p1.json is PartyC(P2) while NAME of P2 in p2.json is (P2).
The three parties follow the steps below to install rosetta-io and generate program check_config_json.
#install_path
$ git clone --recurse https://github.com/LatticeX-Foundation/Rosetta-IO.git
$ cd Rosetta-IO
$ mkdir -p build
$ cd build
$ export install_path=~/.local/rosetta-io
$ cmake ../ -DCMAKE_INSTALL_PREFIX=${install_path}
$ core_num=$(nproc)
$ make -j${core_num} && make install
$ cd ../example
$ g++ check_config_json.cpp -o check_config_json -I${install_path}/include -L${install_path}/lib -lio -Wl,-rpath=${install_path}/lib
Next, the three parties start their program respectively. Here is what P0 does.
$ ./check_config_json p0p1.json P0
Here is what P1 does.
$ ./check_config_json p0p1.json P1
Here is what P2 does.
$ ./check_config_json p2.json P2
The three parties get different results from their console terminals. Here is what P0 gets.
send data to P1
send data to P2
recv data from P1, size:716
recv data from P2, size:710
Here is what P1 gets.
send data to P0
send data to P2
recv data from P0, size:716
recv data from P2, size:710
Here is what P2 gets
send data to P0
send data to P1
recv data from P0, size:716
recv data from P1, size:716