基于序列划分策略的Hmmsearch程序两级并行实现

IMPLEMENTATION OF TWO-LEVEL PARALLELIZATION OF HMMSEARCH BASED ON SEQUENCE PARTITION STRATEGY

Hmm search程序是生物序列分析软件包HMMER中最主要程序之一,采用集中式动态负载平衡方案。为了提高其可扩展性,提出一种基于序列划分策略的两级并行算法,采用分布式动态负载平衡方案。通过增加中间层形成“M aster-subM aster-Slave”结构,来改善原“M aster-Slave”结构的不足。实验数据表明改进算法比原算法具有更好的可扩展并行性能。

Hmmsearch is one of the major program of HMMER, which is a biological sequence analysis software packet. And it takes a centralized dynamic load balancing scheme. To improve the Hmmsearch＇s parallel scalability, A new two-level parallel algorithm based on sequence partition strategy is presented, which is a distributed dynamic load balancing scheme. By introducing the intermediate layer, a new Master-subMaster- Slave structure is better than the Master- Slave structure in parallel scalability,as is shown by the tests.