We continue to discuss ZooKeeper which is a wait free coordination for Internet scale systems. It allows distributed processes to co-ordinate with each other through a shared hierarchical namespace of data registers. It is also used for distributed cluster management. It can be used for distributed synchronization in the form of locks, barriers and queues.The Zookeper service is replicated over a set of machines. All machines store a copy of the data (in memory). A leader is elected on service startup. Clients only connect to a single zookeeper server and maintains a TCP connection. Clients can only read from any Zookeeper server, writes go through the leader and need majority consensus. The data model consists of a hierarchical namespace where each node in the namespace is called a ZNode. Every such node in a Zookeeper's namespace is identified by a path. ZNode paths are canonical, absolute and slash-separted and they have no relative references. Paths are appended with a monotonically increasing number. ZNode with the path ending in the lowest number is elected the leader. Each node supports clients to set watch operations which the zookeeper uses to send notifications to the client. All watches are time-ordered but client should handle latency. APIs supported by all nodes are both synchronous and asynchronous
ZooKeeper introduces elements from group messaging, shared registers, and distributed lock services in a replicated centralized service.
#codingexercise
Yesterday we discussed an optimal solution for this problem: Given an integer array, find the maximum for each and every contiguous subarray of size k
Today we see a simple solution which works for small k
void PrintMaxInWindowSizeK(List<int> A, int k)
{
for (int i = 0; i <= A.Count - k; i++)
{
   int max = A[i];
   for (j = 1; j < k ; j++)
   {
       if (A[i+j] > max)
            max = A[i+j];
   }
   Console.WriteLine("{0},", max);
}
}
The same works with a deque of size k by making only one pass over the list of integers.