def failure_graph(graph_object,percentage_of_nodes_of_failure):
  total_nodes=graph_object.number_of_nodes()
  number_of_nodes_to_delete=int((percentage_of_nodes_of_failure*total_nodes)/100)
  nodes=list(graph_object.nodes)
  deleted_nodes=0
  plot_x=[]
  plot_y_pl=[]
  plot_y_lc=[]
  plot_y_aic=[]
  invl=(total_nodes*(percentage_of_nodes_of_failure/100))//20
  for i in range(number_of_nodes_to_delete):
    node_to_delete=random.choice(nodes)
    nodes.remove(node_to_delete)
    graph_object.remove_node(node_to_delete)
    deleted_nodes+=1
    if(deleted_nodes%invl==0):
      plot_x.append(deleted_nodes/total_nodes)
      #average path length
      apl=[]
      compo=0
      for C in (G.subgraph(c).copy() for c in nx.connected_components(graph_object)):
        compo+=1
        apl.append(nx.average_shortest_path_length(C))
      plot_y_pl.append(max(apl))
      #relative size of largest Cluster
      largest_cc = max(nx.connected_components(graph_object), key=len)
      plot_y_lc.append(len(largest_cc)/graph_object.number_of_nodes())
      #average size of isolated Clusters
      clusters=list(nx.connected_components(G))
      icl=[]
      for clu in clusters:
        icl.append(len(clu))
      plot_y_aic.append(np.mean(icl))
  return plot_x,plot_y_pl,plot_y_lc,plot_y_aic