Depth first search in a distributed way

问题:

I have tried implementing depth first search in c# but I am not exactly sure how to do this the distributed computing way. If you guys can help me out in this i would be really grateful :) You can find my DFS code below

public class DFS
{ 
static List<string> traversedList = new List<string>();
static List<string> parentList = new List<string>();

public static void Main(string[] args)
{

    int N = 100;
    int M = N * 4;
    int P = N * 16;

    Stack newstack = new Stack();

    List<string> global_list=new List<string>();

    StreamReader file = new StreamReader("my input file");

    string text = file.ReadToEnd();

    string[] lines = text.Split('\n');

    string[][] array1 = new string[lines.Length][];

    for (int i = 0; i < lines.Length; i++)
    {
        lines[i] = lines[i].Trim();
        string[] words = lines[i].Split(' ');

        array1[i] = new string[words.Length];

        for (int j = 0; j < words.Length; j++)
        {
            array1[i][j] = words[j];
        }
    }

    StreamWriter sr = new StreamWriter(args[0]);

    for (int i = 0; i < array1.Length; i++)
    {
        for (int j = 0; j < array1[i].Length; j++)
        {
            if (j != 0 )
            {
                sr.Write(array1[i][0] + ":" + array1[i][j]);
                Console.WriteLine(array1[i][0] + ":" + array1[i][j]);
                sr.Write(sr.NewLine);
            }
        }

    }

    int start_no = Convert.ToInt32(args[args.Length-1]);

    traversedList.Add(start_no.ToString());
    parentList.Add("root");
    dfs(array1, start_no);

    for (int z = 0; z < traversedList.Count; z++)
    {
            Console.WriteLine(traversedList.ElementAt(z) + " "+parentList.ElementAt(z)+" "+(z+1));
     }
    Console.ReadLine();
}

private static void dfs(string[][] array, int point)
{
    for (int z = 1; z < array[point].Length; z++)
        {
            if ((!traversedList.Contains(array[point][z])))
            {
                traversedList.Add(array[point][z]);
                parentList.Add(point.ToString());
                dfs(array, int.Parse(array[point][z]));
            }
        }
        return;
}   

}

回答1:

You should read the literature from the computer chess (now evolved into world-champion computerized game players) world. They've been doing distributed depth-first search for some 30 years, and have lots of ideas. It is tricky to get right, because you have to distribute the work evenly without knowing how much work a particular branch might contain.

Check out Monty Newborn or McGill University, which seemed to be quite a hotbed of this some 10 years ago.

And of course, GIYF: "distributed depth first search" produced a reference to this paper: Distributed Algorithms for Depth-First Search. I'd guess it contains lots of ideas from the computer chess world.

The problems for *shared memory" DFS a bit less difficult; you don't have to send messages to your distributed helpers, you can simply pass a pointer :-} It helps to have a language which provides you with parallelism and manages explosive parallelism growth that can occur if your program forks on every branch. I offer an example 4x4 N-puzzle solver built in a parallel programming language I designed. (This example was one of my earliest SO posts!).