me/p2p-rendering-computation
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
nix
p2p-rendering-computation/Docs/DocsDeprecated/docs.org

24 KiB
Raw Permalink Blame History

yes* Chapter 1: Introduction

Abstract

This project focuses on creating a framework on running heavy tasks that a regular computer cannot run easily such as graphically demanding video games, rendering 3D animations , protein folding simulations. In this project the major focus will not be on the financial incentive part. A peer to peer network will be created to help run tasks decentrally, increasing bandwidth for running tasks. To ensure the tasks in the peer to peer network do not corrupt the server 0S (Operating System), they will be executed in a virtual environment in the server.

The main aim of this project was to create a custom peer to peer network. The user acting as the client has total flexibility on how to batch the tasks and the user acting as the server has complete flexibility on tracking the container's usages and killing the containers at any point of time.

Motivation

Many of the users rely on our PC / Laptop or servers that belong to a server farm to run heavy tasks and with the demand of high creativity requires higher computing power. Buying a powerful computer every few years to run a bunch of heavy tasks which are not executed as frequently to reap the benefits can be inefficient utilization of hardware. On the other end, renting servers to run these heavy tasks can be really useful. Ethically speaking this is leading to monopolisation of computing power similar to what is happening in the web server area. By using peer to peer principles it is possible to remove the monopolisation factor and increase the bandwidth between the client and server.

Installation

Over here we will cover the basic steps to get the server and client side running.

Latest release install

https://github.com/Akilan1999/p2p-rendering-computation/releases

Install from Github master branch

Install Go lang

The entire the implementation of this project is done using Go lang. Thus, we need go lang to compile to code to a binary file. Instructions to install Go lang

Install Docker

In this project the choice of virtualization is Docker due to it's wide usage in the developer community. In the server module we use the Docker Go API to create and interact with the containers.

Instructions to install docker

Instructions to install docker GPU

// Do ensure that the docker command does not need sudo to run
sudo chmod 666 /var/run/docker.sock

Build Project and install project

To set up the internal dependencies and build the entire go code into a single binary

make install
For Windows

To set up P2PRC on Windows, simply run this batch file. Make sure you are not in admin mode when running this.

.\install.bat

Add appropriate paths to .bashrc 

export P2PRC=/<PATH>/p2p-rendering-computation
export PATH=/<PATH>/p2p-rendering-computation:${PATH}

Set up configuration file 

make configfile

Open the config file config.json and add the IPv6 address if you have one.

Test if binary works 

p2prc --help
Output: 
NAME:
   p2p-rendering-computation - p2p cli application to create and access VMs in other servers

USAGE:
   p2prc [global options] command [command options] [arguments...]

VERSION:
   <version no>

COMMANDS:
   help, h  Shows a list of commands or help for one command

GLOBAL OPTIONS:
   --Server, -s                           Starts server (default: false) [$SERVER]
   --UpdateServerList, --us               Update List of Server available based on servers iptables (default: false) [$UPDATE_SERVER_LIST]
   --ListServers, --ls                    List servers which can render tasks (default: false) [$LIST_SERVERS]
   --AddServer value, --as value          Adds server IP address to iptables [$ADD_SERVER]
   --ViewImages value, --vi value         View images available on the server IP address [$VIEW_IMAGES]
   --CreateVM value, --touch value        Creates Docker container on the selected server [$CREATE_VM]
   --ContainerName value, --cn value      Specifying the container run on the server side [$CONTAINER_NAME]
   --RemoveVM value, --rm value           Stop and Remove Docker container (IP:port) accompanied by container ID via --ID or --id [$REMOVE_VM]
   --ID value, --id value                 Docker Container ID [$ID]
   --Ports value, -p value                Number of ports to open for the Docker Container [$NUM_PORTS]
   --GPU, --gpu                           Create Docker Containers to access GPU (default: false) [$USE_GPU]
   --Specification value, --specs value   Specs of the server node [$SPECS]
   --SetDefaultConfig, --dc               Sets a default configuration file (default: false) [$SET_DEFAULT_CONFIG]
   --NetworkInterfaces, --ni              Shows the network interface in your computer (default: false) [$NETWORK_INTERFACE]
   --ViewPlugins, --vp                    Shows plugins available to be executed (default: false) [$VIEW_PLUGIN]
   --TrackedContainers, --tc              View (currently running) containers which have been created from the client side  (default: false) [$TRACKED_CONTAINERS]
   --ExecutePlugin value, --plugin value  Plugin which needs to be executed [$EXECUTE_PLUGIN]
   --CreateGroup, --cgroup                Creates a new group (default: false) [$CREATE_GROUP]
   --Group value, --group value           group flag with argument group ID [$GROUP]
   --Groups, --groups                     View all groups (default: false) [$GROUPS]
   --RemoveContainerGroup, --rmcgroup     Remove specific container in the group (default: false) [$REMOVE_CONTAINER_GROUP]
   --RemoveGroup value, --rmgroup value   Removes the entire group [$REMOVE_GROUP]
   --Generate value, --gen value          Generates a new copy of P2PRC which can be modified based on your needs [$GENERATE]
   --ModuleName value, --mod value        New go project module name [$MODULENAME]
   --PullPlugin value, --pp value         Pulls plugin from git repos [$PULLPLUGIN]
   --RemovePlugin value, --rp value       Removes plugin [$REMOVEPLUGIN]
   --help, -h                             show help (default: false)
   --version, -v                          print the version (default: false)

Using basic commands

Start as a server

Do ensure you have Docker installed for this

p2prc -s

View server Specification 

p2prc --specs=<ip address>

Run container

use the --gpu if you know the other machine has a gpu.

p2prc --touch=<server ip address> -p <number of ports> --gpu

Remove container

The docker id is present in the output where you create a container

p2prc --rm=<server ip address> --id=<docker container id>

Adding servers to ip table 

p2prc --as=<server ip address you want to add>

Update ip table 

p2prc --us

List Servers 

p2prc --ls

View Network interfaces 

p2prc --ni

Viewing Containers created Client side 

p2prc --tc

read more on tracking containers

Running plugin 

p2prc --plugin <plugin name> --id <container id or group id>

Create group 

p2prc --cgroup

Add container to group 

p2prc --group <group id> --id <container id>

View groups 

p2prc --groups

View specific group 

p2prc --group <group id>

Delete container from group 

p2prc --rmcgroup --group <group id> --id <container id>

Delete entire group 

p2prc --rmgroup <group id>

read more on grouping containers ### Extending usecase of P2PRC (Requires a go compiler to run)

p2prc --gen <project name> --mod <go module name>

read more about the generate module

Pulling plugin from a remote repo 

p2prc --pp <repo link>

Deleting plugin from the plugin directory 

p2prc --rp <plugin name>

Added custom metadata about the current node 

p2prc --amd "custom metadata"

Using Plugins

This feature is still Under Development: Read more on the implementation

Dependencies
Run Test Cases

  • Generate Test Case Ansible file

    • make testcases
  • Enter inside plugin directory and run tests.

[!NOTE] That docker needs to installed and needs to run without sudo. Refer the section Install Docker. - cd plugin - go test .

P2P Module Implementation

The P2P module (i.e Peer to Peer Module) is responsible for storing the IP table and interacting with the IP table. In the following implementation of the P2P module ,the IP table stores information about servers available in the network. The other functionality the P2P module takes care of is doing the appropriate speed tests to the servers in the IP table. This is for informing the users about nodes which are close by and nodes which have quicker uploads and downloads speeds. The module is responsible to ensure that there are no duplicate server IPs in the IP table and to remove all server IPs which are not pingable.

images/p2pmoduleArch.png
UML diagram of P2P module

The peer to peer implementation was built from scratch. This is because other peer to peer libraries were on the implementation of the Distributed hash table. At the current moment all those heavy features are not needed because the objective is to search and list all possible servers available. The limitation being that to be a part of the network the user has to know at least 1 server. The advantage of building from scratch makes the module super light and possibility for custom functions and structs. The sub topics below will mention the implementations of each functionality in depth.

IP Table

The ip table file is a json as the format with a list of servers ip addresses, latencies, downloads and uploads speeds. The functions implemented include read file, write file and remove duplicate IP addresses. The remove duplicate IP address function exists because sometimes servers IP tables can have the same ip addresses as what the client has. The path of the IP table json file is received from the configuration module.

{
  "ip_address": [
    {
      "ipv4": "<ipv4 address>",
      "latency": "<latency>",
      "download": "<download>",
      "upload": "<upload>"
      "port no": "<server port no>",
    }
  ]
}

Latency

The latency is measured in milliseconds. The route /server_info is called from the server and time it takes to provide a json response is recorded.

NAT Traversal

P2PRC currently supports TURN for NAT traversal.

TURN

The current TURN implementation used is FRP. The TURN server is also required when a P2PRC node is acting as a Server. The TURN server is determined based on the Node with the least amount of latency based on the Nodes available on the IPTable. Once a TURN server is determined there are 2 actions performed. The first one is /FRPPort to the TURN server to receive a port which is used to generate the external port from the TURN server. The flow below describes the workflow.

Client mode

  • Call /FRPPort
http://<turn server ip>:<server port no>/FRPport
  • Call the TURN server in the following manner. The following is a sample code snippet below.
import (
    "github.com/Akilan1999/p2p-rendering-computation/p2p/frp"
)

func main() {
  serverPort, err := frp.GetFRPServerPort("http://" + <lowestLatencyIpAddress.Ipv4> + ":" + lowestLatencyIpAddress.ServerPort)
   if err != nil {
    return nil, err
   }
   // Create 1 second delay to allow FRP server to start
   time.Sleep(1 * time.Second)
   // Starts FRP as a client with
   proxyPort, err := frp.StartFRPClientForServer(<lowestLatencyIpAddress.Ipv4>, serverPort, <the port you want to expose externally>)
   if err != nil {
     return nil, err
   }
}

Language Bindings

Language bindings refers to wrappers to bridge 2 programming languages. This is used in P2PRC to extend calling P2PRC functions in other programming languages. Currently this is done by generating .so and .h from the Go compiler.

How to build shared object files

The easier way 
# Run
make sharedObjects
Or the direct way 
# Run
cd Bindings && go build -buildmode=c-shared -o p2prc.so
If successfully built: 
# Enter into the Bindings directory
cd Bindings
# List files
ls
# Find files
p2prc.h p2prc.so

Workings under the hood

Below are a sample set of commands to open the bindings implementation.

# run
cd Bindings/
# list files
ls 
# search for file
Client.go

In Client go

There a few things to notice which are different from your standard Go programs:

1. We import "C" which means Cgo is required. 
import "C"
2. All functions which are required to be called from other programming languages have comment such as. 
//export <function name>

// ------------ Example ----------------
// The function below allows to externally
// to call the P2PRC function to start containers
// in a specific node in the know list of nodes
// in the p2p network.
// Note: the comment "//export StartContainer".

//export StartContainer
func StartContainer(IP string) (output *C.char) {
     container, err := client.StartContainer(IP, 0, false, "", "")
     if err != nil {
         return C.CString(err.Error())
     }
     return ConvertStructToJSONString(container)
 }
3. While looking through the file (If 2 files are compared it is pretty trivial to notice a common structure). 
// --------- Example ------------

//export StartContainer
func StartContainer(IP string) (output *C.char) {
     container, err := client.StartContainer(IP, 0, false, "", "")
     if err != nil {
         return C.CString(err.Error())
     }
     return ConvertStructToJSONString(container)
}

//export ViewPlugin
func ViewPlugin() (output *C.char) {
    plugins, err := plugin.DetectPlugins()
    if err != nil {
        return C.CString(err.Error())
    }
    return ConvertStructToJSONString(plugins)
}
It is easy to notice that:
  • ConvertStructToJSONString(<go object>): This is a helper function that convert a go object to JSON string initially and converts it to CString.
  • (output *C.char): This is the return type for most of the functions.
A Pseudo code to refer to the common function implementation shape could be represented as: 
func <Function name> (output *C.char) {
      <response>,<error> := <P2PRC function name>(<parameters if needed>)
      if <error> != nil {
          return C.CString(<error>.Error())
      }
      return ConvertStructToJSONString(<response>)
}

Current languages supported

  • Python

Build sample python program

The easier way

# Run
make python
# Expected ouput
Output is in the Directory Bindings/python/export/
# Run
cd Bindings/python/export/
# list files
ls
# Expected output
SharedObjects/  p2prc.py

Above shows a generated folder which consists of a folder called "SharedObjects/" which consists of p2prc.so and p2prc.h files. p2prc.py refers to a sample python script calling P2PRC go functions. To start an any project to extend P2PRC with python, This generated folder can copied and created as a new git repo for P2PRC extensions scripted or used a reference point as proof of concept that P2PRC can be called from other programming languages.

Config Implementation

The configuration module is responsible to store basic information of absolute paths of files being called in the Go code. In a full-fledged Cli the configuration file can be found in the directory etc and from there points to location such as where the IP table file is located. In the future implementation the config file will have information such as number of hops and other parameters to tweak and to improve the effectiveness of the peer to peer network. The configuration module was implemented using the library Viper. The Viper library automates features such as searching in default paths to find out if the configuration file is present. If the configuration file is not present in the default paths then it auto generates the configuration file. The configurations file can be in any format. In this project the configuration file was generated using JSON format.

{
 "MachineName": "pc-74-120.customer.ask4.lan",
 "IPTable": "/Users/akilan/Documents/p2p-rendering-computation/p2p/iptable/ip_table.json",
 "DockerContainers": "/Users/akilan/Documents/p2p-rendering-computation/server/docker/containers/",
 "DefaultDockerFile": "/Users/akilan/Documents/p2p-rendering-computation/server/docker/containers/docker-ubuntu-sshd/",
 "SpeedTestFile": "/Users/akilan/Documents/p2p-rendering-computation/p2p/50.bin",
 "IPV6Address": "",
 "PluginPath": "/Users/akilan/Documents/p2p-rendering-computation/plugin/deploy",
 "TrackContainersPath": "/Users/akilan/Documents/p2p-rendering-computation/client/trackcontainers/trackcontainers.json",
 "ServerPort": "8088",
 "GroupTrackContainersPath": "/Users/akilan/Documents/p2p-rendering-computation/client/trackcontainers/grouptrackcontainers.json",
 "FRPServerPort": "True",
 "BehindNAT": "True",
 "CustomConfig": null
}

Abstractions

The Abstractions package consists of black-boxed functions for P2PRC.

Functions

  • Init(<Project name>): Initializes P2PRC with all the needed configurations.
  • Start(): Starts p2prc as a server and makes it possible to extend by adding other routes and functionality to P2PRC.
  • MapPort(<port no>): On the local machine the port you want to export to world.
  • StartContainer(<ip address>): The machine on the p2p network where you want to spin up a docker container.
  • RemoveContainer(<ip address>,<container id>): Terminate container based on the IP address and container name.
  • GetSpecs(<ip address>): Get specs of a machine on the network based on the IP address.
  • ViewIPTable(): View the IP table which about nodes in the network.
  • UpdateIPTable(): Force update IP table to learn about new nodes faster.

NAT Traversal

P2PRC currently supports TURN for NAT traversal.

TURN

The current TURN implementation used is FRP. The TURN server is also required when a P2PRC node is acting as a Server. The TURN server is determined based on the Node with the least amount of latency based on the Nodes available on the IPTable. Once a TURN server is determined there are 2 actions performed. The first one is /FRPPort to the TURN server to receive a port which is used to generate the external port from the TURN server. The flow below describes the workflow.

Client mode

  • Call /FRPPort
http://<turn server ip>:<server port no>/FRPport
  • Call the TURN server in the following manner. The following is a sample code snippet below.
import (
    "github.com/Akilan1999/p2p-rendering-computation/p2p/frp"
)

func main() {
  serverPort, err := frp.GetFRPServerPort("http://" + <lowestLatencyIpAddress.Ipv4> + ":" + lowestLatencyIpAddress.ServerPort)
   if err != nil {
    return nil, err
   }
   // Create 1 second delay to allow FRP server to start
   time.Sleep(1 * time.Second)
   // Starts FRP as a client with
   proxyPort, err := frp.StartFRPClientForServer(<lowestLatencyIpAddress.Ipv4>, serverPort, <the port you want to expose externally>)
   if err != nil {
     return nil, err
   }
}
Reference in New Issue View Git Blame Copy Permalink