jayanthkmr/api-router-engine

Engine to route APIs based on strategy pattern and dependancy injection

API-Router-Engine

Strategy Pattern & Dependency Injection

• Vendor and RouterEngine Class
• Vendor has state for traffic quota, availability and stats
• Router engine has state of vendor list and route strategy
• Different strategies for routing injected during runtime as method
• Same set of checkpoint and recalculation functions works around those strategies, keeping same behaviour intact but with different intent
• Decorators can be added for checkpointing and book-keeping

Simulation

self.run(): 
  
  self.set_vendor_availability():
    for vendor in vendor_list:
      for each row in opened vendor csv:
        vendor.is_available = row[‘API Available’]
  
  self.cater_request_output():
    for row in opened request time csv:
      ans = self.route(row)
      write ans['Request Index', 'Vendors tried'] in output csv

Strategies

• Dummy Route : Route all request to vendor1 Some requests are never fulfilled, when vendor1 is down
• Simple Route: Route request to vendor1, if it is up, else to vendor2 and then, to vendor3 in order All requests are fulfilled but the request tries the order
• Steady State Traffic Route : Route request based on steady state traffic quota only
• Dynamic Traffic Route: Routing request based on dynamic traffic quota based on failure & comeback

Dynamic Traffic Route Algorithm

def route_dynamic_traffic:
  a = get_index_multinomial_single_roll(traffic_prob_list) 
  for vendor in self.vendors_list[a:]:
    vendor_tried_list.append(vendor.label) 
    self.checkpoint_till_time_sec(vendor, time_sec) 
    if vendor.is_available[time_min - 1]:
      heappush(vendor.stats['success'], time_sec) 
      write_row['Vendors tried'] = '|'.join(vendor_tried_list) 
      break
    else:
      heappush(vendor.stats['failure'], time_sec)
  traffic_prob_list = self.recalculate_traffic_prob(time_sec) 
  return write_row