2種負載均衡演算法

介面定義:

    public interface ILoadBalance<T>
    {
        T Balance();
    }

實現:

    public class WeightObject<T> where T : class
    {
        int weight;
        T activator;
        public WeightObject(T activator, int weight)
        {
            Activator = activator;
            Weight = weight;
        }
        public int Weight
        {
            get
            {
                return weight;
            }
            private set
            {
                if (value <= 0)
                {
                    throw new ArgumentOutOfRangeException();
                }
                weight = value;
            }
        }

        public T Activator
        {
            get
            {
                return activator;
            }
            private set
            {
                if (value == null)
                {
                    throw new ArgumentNullException();
                }
                activator = value;
            }
        }
    }
    public class OrderLoadBalance<T> : ILoadBalance<T>
    {
        private readonly object syncRoot = new object();
        private int gcd;
        private int currentIndex = -1;
        private int currentWeight = 0;
        private int maxWeight;
        private List<WeightObject<Func<T>>> list = new List<WeightObject<Func<T>>>();

        public OrderLoadBalance(IEnumerable<WeightObject<Func<T>>> weightObjects)
        {
            list.AddRange(weightObjects);
            gcd = GetGCD();
            maxWeight = list.Select(w => w.Weight).Max();
        }

        private int GetGCD()
        {
            int gcd = 1;
            int minWeight = list.Select(w => w.Weight).Min();
            for (int i = 1; i < minWeight; i++)
            {
                bool isFound = true;
                foreach (var weightObject in list)
                {
                    if (weightObject.Weight % i != 0)
                    {
                        isFound = false;
                        break;
                    }
                }
                if (isFound) gcd = i;
            }
            return gcd;
        }

        [MethodImpl(MethodImplOptions.Synchronized)]
        public T Balance()
        {
            lock (syncRoot)
            {
                while (true)
                {
                    currentIndex = (currentIndex + 1) % list.Count;
                    if (0 == currentIndex)
                    {
                        currentWeight = currentWeight - gcd;
                        if (0 >= currentWeight)
                        {
                            currentWeight = maxWeight;
                            if (currentWeight == 0) return list[0].Activator();
                        }
                    }
                    if (list[currentIndex].Weight >= currentWeight)
                    {
                        return list[currentIndex].Activator();
                    }
                }
            }
        }
    }

    public class RandomLoadBalance<T> : ILoadBalance<T>
    {

        private Random random;
        private int totalWeight;
        private List<WeightObject<Func<T>>> list = new List<WeightObject<Func<T>>>();

        public RandomLoadBalance(IEnumerable<WeightObject<Func<T>>> weightObjects)
        {
            list.AddRange(weightObjects);
            totalWeight = list.Select(w => w.Weight).Sum();
            random = new Random();
        }

        public T Balance()
        {
            int r = random.Next(totalWeight) + 1;
                int weight = 0;
                foreach (var item in list)
                {
                    weight += item.Weight;
                    if (weight>=r)
                    {
                        return item.Activator();
                    }
                }
                return list[0].Activator();//
        }
    }

     以上兩種方式均可以實現簡單的均衡演算法,第一種我參考許多前輩的寫法,第二種屬於自己想的.從概率論的角度出發都可以滿足需求,而且兩者效率相當(我原以為第二種實現方式速度更快,很遺憾的是結果證明並非如此,可能是Random對象取隨機數比較慢,我理論上認為沒有鎖會更快的),我個人覺得方法二更好,理由是離散型更好,方法一雖然概率上不錯,但是會連續密集的訪問同一對象.作為一種均衡演算法我覺得還是離散性高比較好,因為這樣可以更好的錯開密集訪問!!!