Performance enhancement of smith-waterman algorithm using hybrid model: Comparing the MPI and hybrid programming paradigm on SMP clusters

Abstract

Nowadays, database pattern searching is the most heavily used operation in computational biology. Indeed, sequence alignment algorithm plays an important role to find the homologous groups of sequences which may help to determine the function of new sequences. Meanwhile Smith-Waterman algorithm is one of the most prominent pattern matching algorithms. However, it cost the large quantity of time and resource power. By the aid of parallel hardware and software architecture it becomes more feasible to get the fast and accurate result in efficient time. In this paper, Smith-Waterman algorithm is parallelized base on various types of parallel programming, pure MPI, pure OpenMP and Hybrid MPI-OpenMP model. In addition, based on the experiments it will be proved that hybrid programming which employ the coarse grain and fine grain parallelization, is more efficient compare with pure MPI and pure OpenMP in cluster of SMP machines.

Parallel Biological Sequence Comparison on Heterogeneous High Performance Computing Platforms with BSP++

“…In [1], a cluster of PS3s is used where MPI is used for communication among the PS3. All the cluster approaches use MPI for communication and [15] uses MPI and OpenMP. For each we list GCUPS and, as an indication, the speed-up of such methods.…”

Shared Memory and Hardware Utilizations for the Parallelization of Local Sequences Alignment using SW Algorithm: A Review

“…Most current discussions in local sequence alignment focus on multicore architectures with shared memory such as divide and conquer techniques [16][17][18], striped SW [19][20][21][22], Instruction-set Processor (ASIP) architecture [23], data compression [24,25], genome assembly (re-sequence) algorithms [26,27], and Symmetric Multi-Processing (SMP) architecture [28,29]. This section discusses in detail these algorithms along with their advantage and disadvantage.…”

“…Numerous experiments have established test-bed for comparing three models to parallelize the SW algorithm including pure MPI, pure OpenMP, and a hybrid model using MPI/OpenMP [28,29]. Furthermore, evaluations and measurements for the performance of a hybrid model are tested.…”

Exact Parallel Alignment of Megabase Genomic Sequences With Tunable Work Distribution

“…Therefore, even though a high amount of comparisons is made, the size of the resultant similarity matrices is not big. For instance, in [20], the maximum size of the query sequence is 2KBP. …”

“…The longest sequence size compared with the algorithms discussed in this section is 1.1 MBP. The problem treated by [10,11,20] is the comparison of a small query sequence with a genomic database. These databases are often composed by a great number of sequences and each processor compares the same sequence with a subset of the sequences that belong to the database.…”