Title :
Load balancing using past information of queue
Author :
Guha, Dibyajyoti ; Pathak, S.S.
Author_Institution :
G.S.Sanyal Sch. of Telecommun., Indian Inst. of Technol. Kharagpur, Kharagpur, India
Abstract :
We provide an approach to balancing the load and splitting the traffic during the overload condition based on queue occupation using old information of the output interface queue. A common storage space known as bulletin board is used to store the queue length of the output interface buffer. Bulletin board is updated on a regular interval basis and new arrival verifies the queue length of the bulletin board instead of the output interface buffer. The assumption of our model are: packets arrive as a Poisson stream of rate λ1 for data and λ2 for voice. Packets are served according to FIFO and the service time for a packet is exponentially distributed with mean 1. The Partial Buffer Sharing scheme is applied to partition the output buffer between voice and data packets. The growth of voice buffer is bounded so that queueing delay is restricted for voice packets. Data packet can take their space if the allocated voice packet buffer is unused on an arrival of data packet. But voice packets are prohibited to take the space of unused allocated data packets. The bulletin board fetches the output queue length evolution on a regular interval basis. The model provides an abstraction of simple load balancing scheme in setting when to split the traffic and send it over another path. This model estimates the bandwidth usage on a local granularity basis instead of running periodic probe messages to estimate the bandwidth between two nodes.
Keywords :
Internet telephony; bandwidth allocation; delays; queueing theory; resource allocation; stochastic processes; telecommunication traffic; voice mail; Poisson stream; bandwidth usage; bulletin board; common storage space; data packets; load balancing; output interface buffer; output interface queue; overload condition; partial buffer sharing scheme; queue past information; queueing delay; traffic splitting; voice packet buffer allocation; Bandwidth; Load management; Markov processes; Probability density function; Queueing analysis; Web servers;
Conference_Titel :
Recent Advances in Intelligent Computational Systems (RAICS), 2011 IEEE
Conference_Location :
Trivandrum
Print_ISBN :
978-1-4244-9478-1
DOI :
10.1109/RAICS.2011.6069412