Quality of service on services selection using anycast techniques: A convergence analysis

Anycast is an effective technique to select system servers among many and by spreading client requests achieve high performance and scalability. In our previous work, we presented the Global Application Layer Anycast (GALA) system. By combining real time network distances measures and geolocation, it performed better than its inherited algorithm, GAA. In this work we analyze how fast the GALA algorithm can find the best server to attend a client request and comparing its convergence to GAA. Using simulations, we show that it converges much faster and we propose a maximum selection time metric to be used in the selection process. Experimental results reveal that the GALA algorithm is two times better than the GAA considering the metric proposed.