A Decentralized Method for Scaling Up Genome Similarity Search Services

As genome sequence databases grow in size, the accuracy and speed of sequence similarity detection become more important. There is an increasing number of methods being used for detecting sequence similarity. Meanwhile the demands for genome sequence search and alignment services are also increasing. It is a challenge to scale up the computer systems for hosting various methods and serving requests to these methods in a timely manner. Traditional clusters, which are used in most of scientific centers, can not cope with this challenge. This paper tackles this problem in a novel way, which treats the sequence search requests as content requests to both genome databases and similarity detection methods; therefore, scaling up the computer systems that serve these contents is a process of constructing content distribution network. The paper gives a decentralized method to dynamically construct content distribution networks for a variety of genome sequence similarity detection services. It also provides a scheduling algorithm for efficiently using content nodes. Our simulation study shows that scalability and high content node utilization can be achieved in such a system while the cost of achieving remains reasonable.